/*
* changed so that it confirms to ext2fs_check_dir_entry
*/
static int
ext2fs_dirbadentry(struct vnode *dp, struct ext2fs_direct *de,
int entryoffsetinblock)
{
struct ufsmount *ump = VFSTOUFS(dp->v_mount);
int dirblksiz = ump->um_dirblksiz;
const char *error_msg = NULL;
int reclen = fs2h16(de->e2d_reclen);
int namlen = de->e2d_namlen;
if (reclen < EXT2FS_DIRSIZ(1)) /* e2d_namlen = 1 */
error_msg = "rec_len is smaller than minimal";
else if (reclen % 4 != 0)
error_msg = "rec_len % 4 != 0";
else if (namlen > EXT2FS_MAXNAMLEN)
error_msg = "namlen > EXT2FS_MAXNAMLEN";
else if (reclen < EXT2FS_DIRSIZ(namlen))
error_msg = "reclen is too small for name_len";
else if (entryoffsetinblock + reclen > dirblksiz)
error_msg = "directory entry across blocks";
else if (fs2h32(de->e2d_ino) >
VTOI(dp)->i_e2fs->e2fs.e2fs_icount)
error_msg = "inode out of bounds";
if (error_msg != NULL) {
printf( "bad directory entry: %s\n"
"offset=%d, inode=%lu, rec_len=%d, name_len=%d \n",
error_msg, entryoffsetinblock,
(unsigned long) fs2h32(de->e2d_ino),
reclen, namlen);
panic("ext2fs_dirbadentry");
}
return error_msg == NULL ? 0 : 1;
}
/*
* Attempt to expand the size of a directory
*/
static int
expanddir(struct ext2fs_dinode *dp, char *name)
{
daddr_t lastbn, newblk;
struct bufarea *bp;
char *firstblk;
lastbn = lblkno(&sblock, inosize(dp));
if (lastbn >= NDADDR - 1 || fs2h32(dp->e2di_blocks[lastbn]) == 0 ||
inosize(dp) == 0) {
return (0);
}
if ((newblk = allocblk()) == 0) {
return (0);
}
dp->e2di_blocks[lastbn + 1] = dp->e2di_blocks[lastbn];
dp->e2di_blocks[lastbn] = h2fs32(newblk);
inossize(dp, inosize(dp) + sblock.e2fs_bsize);
dp->e2di_nblock = h2fs32(fs2h32(dp->e2di_nblock) + 1);
bp = getdirblk(fs2h32(dp->e2di_blocks[lastbn + 1]),
sblock.e2fs_bsize);
if (bp->b_errs)
goto bad;
if ((firstblk = malloc(sblock.e2fs_bsize)) == NULL)
err(8, "cannot allocate first block");
memcpy(firstblk, bp->b_un.b_buf, sblock.e2fs_bsize);
bp = getdirblk(newblk, sblock.e2fs_bsize);
if (bp->b_errs) {
free(firstblk);
goto bad;
}
memcpy(bp->b_un.b_buf, firstblk, sblock.e2fs_bsize);
free(firstblk);
dirty(bp);
bp = getdirblk(fs2h32(dp->e2di_blocks[lastbn + 1]),
sblock.e2fs_bsize);
if (bp->b_errs)
goto bad;
emptydir.dot_reclen = h2fs16(sblock.e2fs_bsize);
memcpy(bp->b_un.b_buf, &emptydir, sizeof emptydir);
pwarn("NO SPACE LEFT IN %s", name);
if (preen)
printf(" (EXPANDED)\n");
else if (reply("EXPAND") == 0)
goto bad;
dirty(bp);
inodirty();
return (1);
bad:
dp->e2di_blocks[lastbn] = dp->e2di_blocks[lastbn + 1];
dp->e2di_blocks[lastbn + 1] = 0;
inossize(dp, inosize(dp) - sblock.e2fs_bsize);
dp->e2di_nblock = h2fs32(fs2h32(dp->e2di_nblock) - 1);
freeblk(newblk);
return (0);
}
__compactcall void
ext2fs_ls(struct open_file *f, const char *pattern)
{
struct file *fp = (struct file *)f->f_fsdata;
size_t block_size = fp->f_fs->e2fs_bsize;
char *buf;
size_t buf_size;
lsentry_t *names = NULL;
fp->f_seekp = 0;
while (fp->f_seekp < (off_t)fp->f_di.e2di_size) {
struct ext2fs_direct *dp, *edp;
int rc = buf_read_file(f, &buf, &buf_size);
if (rc)
goto out;
if (buf_size != block_size || buf_size == 0)
goto out;
dp = (struct ext2fs_direct *)buf;
edp = (struct ext2fs_direct *)(buf + buf_size);
for (; dp < edp;
dp = (void *)((char *)dp + fs2h16(dp->e2d_reclen))) {
const char *t;
if (fs2h16(dp->e2d_reclen) <= 0)
goto out;
if (fs2h32(dp->e2d_ino) == 0)
continue;
if (dp->e2d_type >= NELEM(typestr) ||
!(t = typestr[dp->e2d_type])) {
/*
* This does not handle "old"
* filesystems properly. On little
* endian machines, we get a bogus
* type name if the namlen matches a
* valid type identifier. We could
* check if we read namlen "0" and
* handle this case specially, if
* there were a pressing need...
*/
printf("bad dir entry\n");
goto out;
}
lsadd(&names, pattern, dp->e2d_name,
strlen(dp->e2d_name), fs2h32(dp->e2d_ino), t);
}
fp->f_seekp += buf_size;
}
lsprint(names);
out: lsfree(names);
}
u_int64_t
inosize(struct ext2fs_dinode *dp)
{
u_int64_t size = fs2h32(dp->e2di_size);
if ((fs2h16(dp->e2di_mode) & IFMT) == IFREG)
size |= (u_int64_t)fs2h32(dp->e2di_size_high) << 32;
if (size > INT32_MAX)
(void)setlarge();
return size;
}
/*
* Search a directory for a name and return its
* inode number.
*/
static int
search_directory(const char *name, int length, struct open_file *f,
ino32_t *inumber_p)
{
struct file *fp = (struct file *)f->f_fsdata;
struct mfs_sblock *fs = fp->f_fs;
struct mfs_direct *dp;
struct mfs_direct *dbuf;
size_t buf_size;
int namlen;
int rc;
fp->f_seekp = 0;
while (fp->f_seekp < (off_t)fp->f_di.mdi_size) {
rc = buf_read_file(f, (void *)&dbuf, &buf_size);
if (rc)
return rc;
if (buf_size == 0)
return EIO;
/* XXX we assume, that buf_read_file reads an fs block and
* doesn't truncate buffer. Currently i_size in MFS doesn't
* the same as size of allocated blocks, it makes buf_read_file
* to truncate buf_size.
*/
if (buf_size < fs->mfs_block_size)
buf_size = fs->mfs_block_size;
for (dp = dbuf; dp < &dbuf[NR_DIR_ENTRIES(fs)]; dp++) {
char *cp;
if (fs2h32(dp->mfsd_ino) == (ino32_t) 0)
continue;
/* Compute the length of the name */
cp = memchr(dp->mfsd_name, '\0', sizeof(dp->mfsd_name));
if (cp == NULL)
namlen = sizeof(dp->mfsd_name);
else
namlen = cp - (dp->mfsd_name);
if (namlen == length &&
!memcmp(name, dp->mfsd_name, length)) {
/* found entry */
*inumber_p = fs2h32(dp->mfsd_ino);
return 0;
}
}
fp->f_seekp += buf_size;
}
return ENOENT;
}
/*
* Select the desired position for the next block in a file. The file is
* logically divided into sections. The first section is composed of the
* direct blocks. Each additional section contains fs_maxbpg blocks.
*
* If no blocks have been allocated in the first section, the policy is to
* request a block in the same cylinder group as the inode that describes
* the file. Otherwise, the policy is to try to allocate the blocks
* contigously. The two fields of the ext2 inode extension (see
* ufs/ufs/inode.h) help this.
*/
daddr_t
ext2fs_blkpref(struct inode *ip, int32_t lbn, int indx, int32_t *bap)
{
struct m_ext2fs *fs;
int cg, i;
fs = ip->i_e2fs;
/*
* if we are doing contigous lbn allocation, try to alloc blocks
* contigously on disk
*/
if ( ip->i_e2fs_last_blk && lbn == ip->i_e2fs_last_lblk + 1) {
return ip->i_e2fs_last_blk + 1;
}
/*
* bap, if provided, gives us a list of blocks to which we want to
* stay close
*/
if (bap) {
for (i = indx; i >= 0 ; i--) {
if (bap[i]) {
return fs2h32(bap[i]) + 1;
}
}
}
/* fall back to the first block of the cylinder containing the inode */
cg = ino_to_cg(fs, ip->i_number);
return fs->e2fs.e2fs_bpg * cg + fs->e2fs.e2fs_first_dblock + 1;
}
/*
* the problem that is tackled below is the fact that FFS
* includes the terminating zero on disk while EXT2FS doesn't
* this implies that we need to introduce some padding.
* For instance, a filename "sbin" has normally a reclen 12
* in EXT2, but 16 in FFS.
* This reminds me of that Pepsi commercial: 'Kid saved a lousy nine cents...'
* If it wasn't for that, the complete ufs code for directories would
* have worked w/o changes (except for the difference in DIRBLKSIZ)
*/
static void
ext2fs_dirconv2ffs(struct ext2fs_direct *e2dir, struct dirent *ffsdir)
{
memset(ffsdir, 0, sizeof(struct dirent));
ffsdir->d_fileno = fs2h32(e2dir->e2d_ino);
ffsdir->d_namlen = e2dir->e2d_namlen;
ffsdir->d_type = DT_UNKNOWN; /* don't know more here */
#ifdef DIAGNOSTIC
#if MAXNAMLEN < E2FS_MAXNAMLEN
/*
* we should handle this more gracefully !
*/
if (e2dir->e2d_namlen > MAXNAMLEN)
panic("ext2fs: e2dir->e2d_namlen");
#endif
#endif
strncpy(ffsdir->d_name, e2dir->e2d_name, ffsdir->d_namlen);
/* Godmar thinks: since e2dir->e2d_reclen can be big and means
nothing anyway, we compute our own reclen according to what
we think is right
*/
ffsdir->d_reclen = _DIRENT_SIZE(ffsdir);
}
/*
* Verify that a directory entry is valid.
* This is a superset of the checks made in the kernel.
*/
int
dircheck(struct inodesc *idesc, struct ext2fs_direct *dp)
{
int size;
char *cp;
int spaceleft;
u_int16_t reclen = fs2h16(dp->e2d_reclen);
spaceleft = sblock.e2fs_bsize - (idesc->id_loc % sblock.e2fs_bsize);
if (fs2h32(dp->e2d_ino) > maxino ||
reclen == 0 ||
reclen > spaceleft ||
(reclen & 0x3) != 0)
return (0);
if (dp->e2d_ino == 0)
return (1);
if (sblock.e2fs.e2fs_rev < E2FS_REV1 ||
(sblock.e2fs.e2fs_features_incompat & EXT2F_INCOMPAT_FTYPE) == 0)
if (dp->e2d_type != 0)
return (1);
size = EXT2FS_DIRSIZ(dp->e2d_namlen);
if (reclen < size ||
idesc->id_filesize < size /* ||
dp->e2d_namlen > EXT2FS_MAXNAMLEN */)
return (0);
for (cp = dp->e2d_name, size = 0; size < dp->e2d_namlen; size++)
if (*cp == '\0' || (*cp++ == '/'))
return (0);
return (1);
}
/*
* the problem that is tackled below is the fact that FFS
* includes the terminating zero on disk while EXT2FS doesn't
* this implies that we need to introduce some padding.
* For instance, a filename "sbin" has normally a reclen 12
* in EXT2, but 16 in FFS.
* This reminds me of that Pepsi commercial: 'Kid saved a lousy nine cents...'
* If it wasn't for that, the complete ufs code for directories would
* have worked w/o changes (except for the difference in DIRBLKSIZ)
*/
static void
ext2fs_dirconv2ffs(struct ext2fs_direct *e2dir, struct dirent *ffsdir)
{
memset(ffsdir, 0, sizeof(struct dirent));
ffsdir->d_fileno = fs2h32(e2dir->e2d_ino);
ffsdir->d_namlen = e2dir->e2d_namlen;
ffsdir->d_type = DT_UNKNOWN; /* don't know more here */
#ifdef DIAGNOSTIC
/*
* XXX Rigth now this can't happen, but if one day
* MAXNAMLEN != E2FS_MAXNAMLEN we should handle this more gracefully !
*/
/* XXX: e2d_namlen is to small for such comparison
if (e2dir->e2d_namlen > MAXNAMLEN)
panic("ext2fs: e2dir->e2d_namlen");
*/
#endif
strncpy(ffsdir->d_name, e2dir->e2d_name, ffsdir->d_namlen);
/* Godmar thinks: since e2dir->e2d_reclen can be big and means
nothing anyway, we compute our own reclen according to what
we think is right
*/
ffsdir->d_reclen = DIRENT_SIZE(ffsdir);
}
/*
* allocate a new directory
*/
int
allocdir(ino_t parent, ino_t request, int mode)
{
ino_t ino;
struct ext2fs_dinode *dp;
struct bufarea *bp;
struct ext2fs_dirtemplate *dirp;
ino = allocino(request, IFDIR|mode);
dirhead.dot_reclen = h2fs16(12); /* XXX */
dirhead.dotdot_reclen = h2fs16(sblock.e2fs_bsize - 12); /* XXX */
dirhead.dot_namlen = 1;
if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
(sblock.e2fs.e2fs_features_incompat & EXT2F_INCOMPAT_FTYPE))
dirhead.dot_type = EXT2_FT_DIR;
else
dirhead.dot_type = 0;
dirhead.dotdot_namlen = 2;
if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
(sblock.e2fs.e2fs_features_incompat & EXT2F_INCOMPAT_FTYPE))
dirhead.dotdot_type = EXT2_FT_DIR;
else
dirhead.dotdot_type = 0;
dirp = &dirhead;
dirp->dot_ino = h2fs32(ino);
dirp->dotdot_ino = h2fs32(parent);
dp = ginode(ino);
bp = getdirblk(fs2h32(dp->e2di_blocks[0]), sblock.e2fs_bsize);
if (bp->b_errs) {
freeino(ino);
return (0);
}
memcpy(bp->b_un.b_buf, dirp, sizeof(struct ext2fs_dirtemplate));
dirty(bp);
dp->e2di_nlink = h2fs16(2);
inodirty();
if (ino == EXT2_ROOTINO) {
lncntp[ino] = fs2h16(dp->e2di_nlink);
cacheino(dp, ino);
return(ino);
}
if (statemap[parent] != DSTATE && statemap[parent] != DFOUND) {
freeino(ino);
return (0);
}
cacheino(dp, ino);
statemap[ino] = statemap[parent];
if (statemap[ino] == DSTATE) {
lncntp[ino] = fs2h16(dp->e2di_nlink);
lncntp[parent]++;
}
dp = ginode(parent);
dp->e2di_nlink = h2fs16(fs2h16(dp->e2di_nlink) + 1);
inodirty();
return (ino);
}
/*
* Search a directory for a name and return its
* inode number.
*/
static int
search_directory(const char *name, int length, struct open_file *f,
ino32_t *inumber_p)
{
struct file *fp = (struct file *)f->f_fsdata;
struct ext2fs_direct *dp;
struct ext2fs_direct *edp;
char *buf;
size_t buf_size;
int namlen;
int rc;
fp->f_seekp = 0;
/* XXX should handle LARGEFILE */
while (fp->f_seekp < (off_t)fp->f_di.e2di_size) {
rc = buf_read_file(f, &buf, &buf_size);
if (rc)
return rc;
dp = (struct ext2fs_direct *)buf;
edp = (struct ext2fs_direct *)(buf + buf_size);
for (; dp < edp;
dp = (void *)((char *)dp + fs2h16(dp->e2d_reclen))) {
if (fs2h16(dp->e2d_reclen) <= 0)
break;
if (fs2h32(dp->e2d_ino) == (ino32_t)0)
continue;
namlen = dp->e2d_namlen;
if (namlen == length &&
!memcmp(name, dp->e2d_name, length)) {
/* found entry */
*inumber_p = fs2h32(dp->e2d_ino);
return 0;
}
}
fp->f_seekp += buf_size;
}
return ENOENT;
}
/* ARGSUSED */
int
ext2fs_getattr(void *v)
{
struct vop_getattr_args /* {
struct vnode *a_vp;
struct vattr *a_vap;
kauth_cred_t a_cred;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct inode *ip = VTOI(vp);
struct vattr *vap = ap->a_vap;
EXT2FS_ITIMES(ip, NULL, NULL, NULL);
/*
* Copy from inode table
*/
vap->va_fsid = ip->i_dev;
vap->va_fileid = ip->i_number;
vap->va_mode = ip->i_e2fs_mode & ALLPERMS;
vap->va_nlink = ip->i_e2fs_nlink;
vap->va_uid = ip->i_uid;
vap->va_gid = ip->i_gid;
vap->va_rdev = (dev_t)fs2h32(ip->i_din.e2fs_din->e2di_rdev);
vap->va_size = vp->v_size;
vap->va_atime.tv_sec = ip->i_e2fs_atime;
vap->va_atime.tv_nsec = 0;
vap->va_mtime.tv_sec = ip->i_e2fs_mtime;
vap->va_mtime.tv_nsec = 0;
vap->va_ctime.tv_sec = ip->i_e2fs_ctime;
vap->va_ctime.tv_nsec = 0;
#ifdef EXT2FS_SYSTEM_FLAGS
vap->va_flags = (ip->i_e2fs_flags & EXT2_APPEND) ? SF_APPEND : 0;
vap->va_flags |= (ip->i_e2fs_flags & EXT2_IMMUTABLE) ? SF_IMMUTABLE : 0;
#else
vap->va_flags = (ip->i_e2fs_flags & EXT2_APPEND) ? UF_APPEND : 0;
vap->va_flags |= (ip->i_e2fs_flags & EXT2_IMMUTABLE) ? UF_IMMUTABLE : 0;
#endif
vap->va_gen = ip->i_e2fs_gen;
/* this doesn't belong here */
if (vp->v_type == VBLK)
vap->va_blocksize = BLKDEV_IOSIZE;
else if (vp->v_type == VCHR)
vap->va_blocksize = MAXBSIZE;
else
vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
vap->va_bytes = dbtob((u_quad_t)ip->i_e2fs_nblock);
vap->va_type = vp->v_type;
vap->va_filerev = ip->i_modrev;
return (0);
}
static int
ext2fs_checksb(struct ext2fs *fs, int ronly)
{
uint32_t u32;
if (fs2h16(fs->e2fs_magic) != E2FS_MAGIC) {
return (EINVAL); /* XXX needs translation */
}
if (fs2h32(fs->e2fs_rev) > E2FS_REV1) {
#ifdef DIAGNOSTIC
printf("ext2fs: unsupported revision number: %x\n",
fs2h32(fs->e2fs_rev));
#endif
return (EINVAL); /* XXX needs translation */
}
if (fs2h32(fs->e2fs_log_bsize) > 2) { /* block size = 1024|2048|4096 */
#ifdef DIAGNOSTIC
printf("ext2fs: bad block size: %d "
"(expected <= 2 for ext2 fs)\n",
fs2h32(fs->e2fs_log_bsize));
#endif
return (EINVAL); /* XXX needs translation */
}
if (fs2h32(fs->e2fs_rev) > E2FS_REV0) {
char buf[256];
if (fs2h32(fs->e2fs_first_ino) != EXT2_FIRSTINO) {
printf("ext2fs: unsupported first inode position\n");
return (EINVAL); /* XXX needs translation */
}
u32 = fs2h32(fs->e2fs_features_incompat) & ~EXT2F_INCOMPAT_SUPP;
if (u32) {
snprintb(buf, sizeof(buf), EXT2F_INCOMPAT_BITS, u32);
printf("ext2fs: unsupported incompat features: %s\n",
buf);
return EINVAL; /* XXX needs translation */
}
u32 = fs2h32(fs->e2fs_features_rocompat) & ~EXT2F_ROCOMPAT_SUPP;
if (!ronly && u32) {
snprintb(buf, sizeof(buf), EXT2F_ROCOMPAT_BITS, u32);
printf("ext2fs: unsupported ro-incompat features: %s\n",
buf);
return EROFS; /* XXX needs translation */
}
}
return (0);
}
/*
* Check if a directory is empty or not.
* Inode supplied must be locked.
*
* Using a struct dirtemplate here is not precisely
* what we want, but better than using a struct ext2fs_direct.
*
* NB: does not handle corrupted directories.
*/
int
ext2fs_dirempty(struct inode *ip, ino_t parentino, kauth_cred_t cred)
{
off_t off;
struct ext2fs_dirtemplate dbuf;
struct ext2fs_direct *dp = (struct ext2fs_direct *)&dbuf;
int error, namlen;
size_t count;
#define MINDIRSIZ (sizeof (struct ext2fs_dirtemplate) / 2)
for (off = 0; off < ext2fs_size(ip); off += fs2h16(dp->e2d_reclen)) {
error = ufs_bufio(UIO_READ, ITOV(ip), (void *)dp, MINDIRSIZ,
off, IO_NODELOCKED, cred, &count, NULL);
/*
* Since we read MINDIRSIZ, residual must
* be 0 unless we're at end of file.
*/
if (error || count != 0)
return (0);
/* avoid infinite loops */
if (dp->e2d_reclen == 0)
return (0);
/* skip empty entries */
if (dp->e2d_ino == 0)
continue;
/* accept only "." and ".." */
namlen = dp->e2d_namlen;
if (namlen > 2)
return (0);
if (dp->e2d_name[0] != '.')
return (0);
/*
* At this point namlen must be 1 or 2.
* 1 implies ".", 2 implies ".." if second
* char is also "."
*/
if (namlen == 1)
continue;
if (dp->e2d_name[1] == '.' && fs2h32(dp->e2d_ino) == parentino)
continue;
return (0);
}
return (1);
}
(void)ext2fs_setsize(ip, 0);
memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
}
*vpp = vp;
return (0);
}
static int
ext2fs_checksb(struct ext2fs *fs, int ronly)
{
if (fs2h16(fs->e2fs_magic) != E2FS_MAGIC) {
return (EINVAL); /* XXX needs translation */
}
if (fs2h32(fs->e2fs_rev) > E2FS_REV1) {
#ifdef DIAGNOSTIC
printf("Ext2 fs: unsupported revision number: %x\n",
fs2h32(fs->e2fs_rev));
#endif
return (EINVAL); /* XXX needs translation */
}
if (fs2h32(fs->e2fs_log_bsize) > 2) { /* block size = 1024|2048|4096 */
#ifdef DIAGNOSTIC
printf("Ext2 fs: bad block size: %d "
"(expected <= 2 for ext2 fs)\n",
fs2h32(fs->e2fs_log_bsize));
#endif
return (EINVAL); /* XXX needs translation */
}
if (fs2h32(fs->e2fs_rev) > E2FS_REV0) {
if (fs2h32(fs->e2fs_first_ino) != EXT2_FIRSTINO) {
printf("Ext2 fs: unsupported first inode position\n");
return (EINVAL); /* XXX needs translation */
}
if (fs2h32(fs->e2fs_features_incompat) &
~EXT2F_INCOMPAT_SUPP) {
printf("Ext2 fs: unsupported optional feature\n");
return (EINVAL); /* XXX needs translation */
/*
* ext2fs_rename_replace_dotdot: Change the target of the `..' entry of
* the directory vp from fdvp to tdvp.
*/
static int
ext2fs_rename_replace_dotdot(struct vnode *vp,
struct vnode *fdvp, struct vnode *tdvp,
kauth_cred_t cred)
{
struct ext2fs_dirtemplate dirbuf;
int error;
/* XXX Does it make sense to do this before the sanity checks below? */
KASSERT(0 < VTOI(fdvp)->i_e2fs_nlink);
VTOI(fdvp)->i_e2fs_nlink--;
VTOI(fdvp)->i_flag |= IN_CHANGE;
error = vn_rdwr(UIO_READ, vp, &dirbuf, sizeof dirbuf, (off_t)0,
UIO_SYSSPACE, IO_NODELOCKED, cred, NULL, NULL);
if (error)
return error;
if (dirbuf.dotdot_namlen != 2 ||
dirbuf.dotdot_name[0] != '.' ||
dirbuf.dotdot_name[1] != '.') {
ufs_dirbad(VTOI(vp), (doff_t)12, "bad `..' entry");
return 0;
}
if (fs2h32(dirbuf.dotdot_ino) != VTOI(fdvp)->i_number) {
ufs_dirbad(VTOI(vp), (doff_t)12,
"`..' does not point at parent");
return 0;
}
dirbuf.dotdot_ino = h2fs32(VTOI(tdvp)->i_number);
/* XXX WTF? Why not check error? */
(void)vn_rdwr(UIO_WRITE, vp, &dirbuf, sizeof dirbuf, (off_t)0,
UIO_SYSSPACE, (IO_NODELOCKED | IO_SYNC), cred, NULL, NULL);
return 0;
}
/*
* ext2fs_read_dotdot: Store in *ino_ret the inode number of the parent
* of the directory vp.
*/
static int
ext2fs_read_dotdot(struct vnode *vp, kauth_cred_t cred, ino_t *ino_ret)
{
struct ext2fs_dirtemplate dirbuf;
int error;
KASSERT(vp != NULL);
KASSERT(ino_ret != NULL);
KASSERT(vp->v_type == VDIR);
error = vn_rdwr(UIO_READ, vp, &dirbuf, sizeof dirbuf, (off_t)0,
UIO_SYSSPACE, IO_NODELOCKED, cred, NULL, NULL);
if (error)
return error;
if (dirbuf.dotdot_namlen != 2 ||
dirbuf.dotdot_name[0] != '.' ||
dirbuf.dotdot_name[1] != '.')
/* XXX Panic? Print warning? */
return ENOTDIR;
*ino_ret = fs2h32(dirbuf.dotdot_ino);
return 0;
}
/*
* Convert a component of a pathname into a pointer to a locked inode.
* This is a very central and rather complicated routine.
* If the file system is not maintained in a strict tree hierarchy,
* this can result in a deadlock situation (see comments in code below).
*
* The cnp->cn_nameiop argument is LOOKUP, CREATE, RENAME, or DELETE depending
* on whether the name is to be looked up, created, renamed, or deleted.
* When CREATE, RENAME, or DELETE is specified, information usable in
* creating, renaming, or deleting a directory entry may be calculated.
* If flag has LOCKPARENT or'ed into it and the target of the pathname
* exists, lookup returns both the target and its parent directory locked.
* When creating or renaming and LOCKPARENT is specified, the target may
* not be ".". When deleting and LOCKPARENT is specified, the target may
* be "."., but the caller must check to ensure it does an vrele and vput
* instead of two vputs.
*
* Overall outline of ext2fs_lookup:
*
* check accessibility of directory
* look for name in cache, if found, then if at end of path
* and deleting or creating, drop it, else return name
* search for name in directory, to found or notfound
* notfound:
* if creating, return locked directory, leaving info on available slots
* else return error
* found:
* if at end of path and deleting, return information to allow delete
* if at end of path and rewriting (RENAME and LOCKPARENT), lock target
* inode and return info to allow rewrite
* if not at end, add name to cache; if at end and neither creating
* nor deleting, add name to cache
*/
int
ext2fs_lookup(void *v)
{
struct vop_lookup_args *ap = v;
struct vnode *vdp; /* vnode for directory being searched */
struct inode *dp; /* inode for directory being searched */
struct buf *bp; /* a buffer of directory entries */
struct ext2fs_direct *ep; /* the current directory entry */
int entryoffsetinblock; /* offset of ep in bp's buffer */
enum {NONE, COMPACT, FOUND} slotstatus;
doff_t slotoffset; /* offset of area with free space */
int slotsize; /* size of area at slotoffset */
int slotfreespace; /* amount of space free in slot */
int slotneeded; /* size of the entry we're seeking */
int numdirpasses; /* strategy for directory search */
doff_t endsearch; /* offset to end directory search */
doff_t prevoff; /* prev entry dp->i_offset */
struct vnode *pdp; /* saved dp during symlink work */
struct vnode *tdp; /* returned by VFS_VGET */
doff_t enduseful; /* pointer past last used dir slot */
u_long bmask; /* block offset mask */
int lockparent; /* 1 => lockparent flag is set */
int wantparent; /* 1 => wantparent or lockparent flag */
int namlen, error;
struct vnode **vpp = ap->a_vpp;
struct componentname *cnp = ap->a_cnp;
struct ucred *cred = cnp->cn_cred;
int flags = cnp->cn_flags;
int nameiop = cnp->cn_nameiop;
struct proc *p = cnp->cn_proc;
int dirblksize = VTOI(ap->a_dvp)->i_e2fs->e2fs_bsize;
bp = NULL;
slotoffset = -1;
*vpp = NULL;
vdp = ap->a_dvp;
dp = VTOI(vdp);
lockparent = flags & LOCKPARENT;
wantparent = flags & (LOCKPARENT|WANTPARENT);
/*
* Check accessiblity of directory.
*/
if ((error = VOP_ACCESS(vdp, VEXEC, cred)) != 0)
return (error);
if ((flags & ISLASTCN) && (vdp->v_mount->mnt_flag & MNT_RDONLY) &&
(cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
return (EROFS);
/*
* We now have a segment name to search for, and a directory to search.
*
* Before tediously performing a linear scan of the directory,
* check the name cache to see if the directory/name pair
* we are looking for is known already.
*/
if ((error = cache_lookup(vdp, vpp, cnp)) >= 0)
return (error);
/*
* Suppress search for slots unless creating
* file and at end of pathname, in which case
* we watch for a place to put the new file in
* case it doesn't already exist.
*/
slotstatus = FOUND;
slotfreespace = slotsize = slotneeded = 0;
if ((nameiop == CREATE || nameiop == RENAME) &&
(flags & ISLASTCN)) {
slotstatus = NONE;
slotneeded = EXT2FS_DIRSIZ(cnp->cn_namelen);
}
/*
* If there is cached information on a previous search of
* this directory, pick up where we last left off.
* We cache only lookups as these are the most common
* and have the greatest payoff. Caching CREATE has little
* benefit as it usually must search the entire directory
* to determine that the entry does not exist. Caching the
* location of the last DELETE or RENAME has not reduced
* profiling time and hence has been removed in the interest
* of simplicity.
*/
bmask = VFSTOUFS(vdp->v_mount)->um_mountp->mnt_stat.f_iosize - 1;
if (nameiop != LOOKUP || dp->i_diroff == 0 ||
dp->i_diroff >ext2fs_size(dp)) {
entryoffsetinblock = 0;
dp->i_offset = 0;
numdirpasses = 1;
} else {
dp->i_offset = dp->i_diroff;
if ((entryoffsetinblock = dp->i_offset & bmask) &&
(error = ext2fs_bufatoff(dp, (off_t)dp->i_offset,
NULL, &bp)))
return (error);
numdirpasses = 2;
}
prevoff = dp->i_offset;
endsearch = roundup(ext2fs_size(dp), dirblksize);
enduseful = 0;
searchloop:
while (dp->i_offset < endsearch) {
/*
* If necessary, get the next directory block.
*/
if ((dp->i_offset & bmask) == 0) {
if (bp != NULL)
brelse(bp);
error = ext2fs_bufatoff(dp, (off_t)dp->i_offset,
NULL, &bp);
if (error != 0)
return (error);
entryoffsetinblock = 0;
}
/*
* If still looking for a slot, and at a dirblksize
* boundary, have to start looking for free space again.
*/
if (slotstatus == NONE &&
(entryoffsetinblock & (dirblksize - 1)) == 0) {
slotoffset = -1;
slotfreespace = 0;
}
/*
* Get pointer to next entry.
* Full validation checks are slow, so we only check
* enough to insure forward progress through the
* directory. Complete checks can be run by patching
* "dirchk" to be true.
*/
ep = (struct ext2fs_direct *)
((char *)bp->b_data + entryoffsetinblock);
if (ep->e2d_reclen == 0 ||
(dirchk &&
ext2fs_dirbadentry(vdp, ep, entryoffsetinblock))) {
int i;
ufs_dirbad(dp, dp->i_offset, "mangled entry");
i = dirblksize -
(entryoffsetinblock & (dirblksize - 1));
dp->i_offset += i;
entryoffsetinblock += i;
continue;
}
/*
* If an appropriate sized slot has not yet been found,
* check to see if one is available. Also accumulate space
* in the current block so that we can determine if
* compaction is viable.
*/
if (slotstatus != FOUND) {
int size = fs2h16(ep->e2d_reclen);
if (ep->e2d_ino != 0)
size -= EXT2FS_DIRSIZ(ep->e2d_namlen);
if (size > 0) {
if (size >= slotneeded) {
slotstatus = FOUND;
slotoffset = dp->i_offset;
slotsize = fs2h16(ep->e2d_reclen);
} else if (slotstatus == NONE) {
slotfreespace += size;
if (slotoffset == -1)
slotoffset = dp->i_offset;
if (slotfreespace >= slotneeded) {
slotstatus = COMPACT;
slotsize = dp->i_offset +
fs2h16(ep->e2d_reclen) - slotoffset;
}
}
}
}
/*
* Check for a name match.
*/
if (ep->e2d_ino) {
namlen = ep->e2d_namlen;
if (namlen == cnp->cn_namelen &&
!memcmp(cnp->cn_nameptr, ep->e2d_name,
(unsigned)namlen)) {
/*
* Save directory entry's inode number and
* reclen in ndp->ni_ufs area, and release
* directory buffer.
*/
dp->i_ino = fs2h32(ep->e2d_ino);
dp->i_reclen = fs2h16(ep->e2d_reclen);
brelse(bp);
goto found;
}
}
prevoff = dp->i_offset;
dp->i_offset += fs2h16(ep->e2d_reclen);
entryoffsetinblock += fs2h16(ep->e2d_reclen);
if (ep->e2d_ino)
enduseful = dp->i_offset;
}
/* notfound: */
/*
* If we started in the middle of the directory and failed
* to find our target, we must check the beginning as well.
*/
if (numdirpasses == 2) {
numdirpasses--;
dp->i_offset = 0;
endsearch = dp->i_diroff;
goto searchloop;
}
if (bp != NULL)
brelse(bp);
/*
* If creating, and at end of pathname and current
* directory has not been removed, then can consider
* allowing file to be created.
*/
if ((nameiop == CREATE || nameiop == RENAME) &&
(flags & ISLASTCN) && dp->i_e2fs_nlink != 0) {
/*
* Creation of files on a read-only mounted file system
* is pointless, so don't proceed any further.
*/
if (vdp->v_mount->mnt_flag & MNT_RDONLY)
return (EROFS);
/*
* Access for write is interpreted as allowing
* creation of files in the directory.
*/
if ((error = VOP_ACCESS(vdp, VWRITE, cred)) != 0)
return (error);
/*
* Return an indication of where the new directory
* entry should be put. If we didn't find a slot,
* then set dp->i_count to 0 indicating
* that the new slot belongs at the end of the
* directory. If we found a slot, then the new entry
* can be put in the range from dp->i_offset to
* dp->i_offset + dp->i_count.
*/
if (slotstatus == NONE) {
dp->i_offset = roundup(ext2fs_size(dp), dirblksize);
dp->i_count = 0;
enduseful = dp->i_offset;
} else {
dp->i_offset = slotoffset;
dp->i_count = slotsize;
if (enduseful < slotoffset + slotsize)
enduseful = slotoffset + slotsize;
}
dp->i_endoff = roundup(enduseful, dirblksize);
dp->i_flag |= IN_CHANGE | IN_UPDATE;
/*
* We return with the directory locked, so that
* the parameters we set up above will still be
* valid if we actually decide to do a direnter().
* We return ni_vp == NULL to indicate that the entry
* does not currently exist; we leave a pointer to
* the (locked) directory inode in ndp->ni_dvp.
* The pathname buffer is saved so that the name
* can be obtained later.
*
* NB - if the directory is unlocked, then this
* information cannot be used.
*/
cnp->cn_flags |= SAVENAME;
if (!lockparent) {
VOP_UNLOCK(vdp, 0);
cnp->cn_flags |= PDIRUNLOCK;
}
return (EJUSTRETURN);
}
/*
* Insert name into cache (as non-existent) if appropriate.
*/
if ((cnp->cn_flags & MAKEENTRY) && nameiop != CREATE)
cache_enter(vdp, *vpp, cnp);
return (ENOENT);
found:
/*
* Check that directory length properly reflects presence
* of this entry.
*/
if (entryoffsetinblock + EXT2FS_DIRSIZ(ep->e2d_namlen)
> ext2fs_size(dp)) {
ufs_dirbad(dp, dp->i_offset, "i_size too small");
error = ext2fs_setsize(dp,
entryoffsetinblock + EXT2FS_DIRSIZ(ep->e2d_namlen));
if (error) {
brelse(bp);
return(error);
}
dp->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* Found component in pathname.
* If the final component of path name, save information
* in the cache as to where the entry was found.
*/
if ((flags & ISLASTCN) && nameiop == LOOKUP)
dp->i_diroff = dp->i_offset &~ (dirblksize - 1);
/*
* If deleting, and at end of pathname, return
* parameters which can be used to remove file.
* If the wantparent flag isn't set, we return only
* the directory (in ndp->ni_dvp), otherwise we go
* on and lock the inode, being careful with ".".
*/
if (nameiop == DELETE && (flags & ISLASTCN)) {
/*
* Write access to directory required to delete files.
*/
if ((error = VOP_ACCESS(vdp, VWRITE, cred)) != 0)
return (error);
/*
* Return pointer to current entry in dp->i_offset,
* and distance past previous entry (if there
* is a previous entry in this block) in dp->i_count.
* Save directory inode pointer in ndp->ni_dvp for dirremove().
*/
if ((dp->i_offset & (dirblksize - 1)) == 0)
dp->i_count = 0;
else
dp->i_count = dp->i_offset - prevoff;
if (dp->i_number == dp->i_ino) {
vref(vdp);
*vpp = vdp;
return (0);
}
if ((error = VFS_VGET(vdp->v_mount, dp->i_ino, &tdp)) != 0)
return (error);
/*
* If directory is "sticky", then user must own
* the directory, or the file in it, else she
* may not delete it (unless she's root). This
* implements append-only directories.
*/
if ((dp->i_e2fs_mode & ISVTX) &&
cred->cr_uid != 0 &&
cred->cr_uid != dp->i_e2fs_uid &&
VTOI(tdp)->i_e2fs_uid != cred->cr_uid) {
vput(tdp);
return (EPERM);
}
*vpp = tdp;
if (!lockparent) {
VOP_UNLOCK(vdp, 0);
cnp->cn_flags |= PDIRUNLOCK;
}
return (0);
}
/*
* If rewriting (RENAME), return the inode and the
* information required to rewrite the present directory
* Must get inode of directory entry to verify it's a
* regular file, or empty directory.
*/
if (nameiop == RENAME && wantparent &&
(flags & ISLASTCN)) {
if ((error = VOP_ACCESS(vdp, VWRITE, cred)) != 0)
return (error);
/*
* Careful about locking second inode.
* This can only occur if the target is ".".
*/
if (dp->i_number == dp->i_ino)
return (EISDIR);
if ((error = VFS_VGET(vdp->v_mount, dp->i_ino, &tdp)) != 0)
return (error);
*vpp = tdp;
cnp->cn_flags |= SAVENAME;
if (!lockparent) {
VOP_UNLOCK(vdp, 0);
cnp->cn_flags |= PDIRUNLOCK;
}
return (0);
}
/*
* Step through the translation in the name. We do not `vput' the
* directory because we may need it again if a symbolic link
* is relative to the current directory. Instead we save it
* unlocked as "pdp". We must get the target inode before unlocking
* the directory to insure that the inode will not be removed
* before we get it. We prevent deadlock by always fetching
* inodes from the root, moving down the directory tree. Thus
* when following backward pointers ".." we must unlock the
* parent directory before getting the requested directory.
* There is a potential race condition here if both the current
* and parent directories are removed before the VFS_VGET for the
* inode associated with ".." returns. We hope that this occurs
* infrequently since we cannot avoid this race condition without
* implementing a sophisticated deadlock detection algorithm.
* Note also that this simple deadlock detection scheme will not
* work if the file system has any hard links other than ".."
* that point backwards in the directory structure.
*/
pdp = vdp;
if (flags & ISDOTDOT) {
VOP_UNLOCK(pdp, 0); /* race to get the inode */
cnp->cn_flags |= PDIRUNLOCK;
if ((error = VFS_VGET(vdp->v_mount, dp->i_ino, &tdp)) != 0) {
if (vn_lock(pdp, LK_EXCLUSIVE | LK_RETRY, p) == 0)
cnp->cn_flags &= ~PDIRUNLOCK;
return (error);
}
if (lockparent && (flags & ISLASTCN)) {
if ((error = vn_lock(pdp, LK_EXCLUSIVE, p)) != 0) {
vput(tdp);
return (error);
}
cnp->cn_flags &= ~PDIRUNLOCK;
}
*vpp = tdp;
} else if (dp->i_number == dp->i_ino) {
vref(vdp); /* we want ourself, ie "." */
*vpp = vdp;
} else {
if ((error = VFS_VGET(vdp->v_mount, dp->i_ino, &tdp)) != 0)
return (error);
if (!lockparent || !(flags & ISLASTCN)) {
VOP_UNLOCK(pdp, 0);
cnp->cn_flags |= PDIRUNLOCK;
}
*vpp = tdp;
}
/*
* Insert name into cache if appropriate.
*/
if (cnp->cn_flags & MAKEENTRY)
cache_enter(vdp, *vpp, cnp);
return (0);
}
/*
* ext2fs_rename_recalculate_fulr: If we have just entered a directory
* into dvp at tulr, and we were about to remove one at fulr for an
* entry named fcnp, fulr may be invalid. So, if necessary,
* recalculate it.
*/
static int
ext2fs_rename_recalculate_fulr(struct vnode *dvp,
struct ufs_lookup_results *fulr, const struct ufs_lookup_results *tulr,
const struct componentname *fcnp)
{
struct mount *mp;
struct ufsmount *ump;
/* XXX int is a silly type for this; blame ufsmount::um_dirblksiz. */
int dirblksiz;
doff_t search_start, search_end;
doff_t offset; /* Offset of entry we're examining. */
struct buf *bp; /* I/O block we're examining. */
char *dirbuf; /* Pointer into directory at search_start. */
struct ext2fs_direct *ep; /* Pointer to the entry we're examining. */
/* XXX direct::d_reclen is 16-bit;
* ufs_lookup_results::ulr_reclen is 32-bit. Blah. */
uint32_t reclen; /* Length of the entry we're examining. */
uint32_t prev_reclen; /* Length of the preceding entry. */
int error;
KASSERT(dvp != NULL);
KASSERT(dvp->v_mount != NULL);
KASSERT(VTOI(dvp) != NULL);
KASSERT(fulr != NULL);
KASSERT(tulr != NULL);
KASSERT(fulr != tulr);
KASSERT(ext2fs_rename_ulr_overlap_p(fulr, tulr));
mp = dvp->v_mount;
ump = VFSTOUFS(mp);
KASSERT(ump != NULL);
KASSERT(ump == VTOI(dvp)->i_ump);
dirblksiz = ump->um_dirblksiz;
KASSERT(0 < dirblksiz);
KASSERT((dirblksiz & (dirblksiz - 1)) == 0);
/* A directory block may not span across multiple I/O blocks. */
KASSERT(dirblksiz <= mp->mnt_stat.f_iosize);
/* Find the bounds of the search. */
search_start = tulr->ulr_offset;
KASSERT(fulr->ulr_reclen < (EXT2FS_MAXDIRSIZE - fulr->ulr_offset));
search_end = (fulr->ulr_offset + fulr->ulr_reclen);
/* Compaction must happen only within a directory block. (*) */
KASSERT(search_start <= search_end);
KASSERT((search_end - (search_start &~ (dirblksiz - 1))) <= dirblksiz);
dirbuf = NULL;
bp = NULL;
error = ext2fs_blkatoff(dvp, (off_t)search_start, &dirbuf, &bp);
if (error)
return error;
KASSERT(dirbuf != NULL);
KASSERT(bp != NULL);
/*
* Guarantee we sha'n't go past the end of the buffer we got.
* dirbuf is bp->b_data + (search_start & (iosize - 1)), and
* the valid range is [bp->b_data, bp->b_data + bp->b_bcount).
*/
KASSERT((search_end - search_start) <=
(bp->b_bcount - (search_start & (mp->mnt_stat.f_iosize - 1))));
prev_reclen = fulr->ulr_count;
offset = search_start;
/*
* Search from search_start to search_end for the entry matching
* fcnp, which must be there because we found it before and it
* should only at most have moved earlier.
*/
for (;;) {
KASSERT(search_start <= offset);
KASSERT(offset < search_end);
/*
* Examine the directory entry at offset.
*/
ep = (struct ext2fs_direct *)
(dirbuf + (offset - search_start));
reclen = fs2h16(ep->e2d_reclen);
if (ep->e2d_ino == 0)
goto next; /* Entry is unused. */
if (fs2h32(ep->e2d_ino) == UFS_WINO)
goto next; /* Entry is whiteout. */
if (fcnp->cn_namelen != ep->e2d_namlen)
goto next; /* Wrong name length. */
if (memcmp(ep->e2d_name, fcnp->cn_nameptr, fcnp->cn_namelen))
goto next; /* Wrong name. */
/* Got it! */
break;
next:
if (! ((reclen < search_end) &&
(offset < (search_end - reclen)))) {
brelse(bp, 0);
return EIO; /* XXX Panic? What? */
}
/* We may not move past the search end. */
KASSERT(reclen < search_end);
KASSERT(offset < (search_end - reclen));
/*
* We may not move across a directory block boundary;
* see (*) above.
*/
KASSERT((offset &~ (dirblksiz - 1)) ==
((offset + reclen) &~ (dirblksiz - 1)));
prev_reclen = reclen;
offset += reclen;
}
/*
* Found the entry. Record where.
*/
fulr->ulr_offset = offset;
fulr->ulr_reclen = reclen;
/*
* Record the preceding record length, but not if we're at the
* start of a directory block.
*/
fulr->ulr_count = ((offset & (dirblksiz - 1))? prev_reclen : 0);
brelse(bp, 0);
return 0;
}
/*
* Convert a component of a pathname into a pointer to a locked inode.
* This is a very central and rather complicated routine.
* If the file system is not maintained in a strict tree hierarchy,
* this can result in a deadlock situation (see comments in code below).
*
* The cnp->cn_nameiop argument is LOOKUP, CREATE, RENAME, or DELETE depending
* on whether the name is to be looked up, created, renamed, or deleted.
* When CREATE, RENAME, or DELETE is specified, information usable in
* creating, renaming, or deleting a directory entry may be calculated.
* If flag has LOCKPARENT or'ed into it and the target of the pathname
* exists, lookup returns both the target and its parent directory locked.
* When creating or renaming and LOCKPARENT is specified, the target may
* not be ".". When deleting and LOCKPARENT is specified, the target may
* be "."., but the caller must check to ensure it does an vrele and vput
* instead of two vputs.
*
* Overall outline of ext2fs_lookup:
*
* check accessibility of directory
* look for name in cache, if found, then if at end of path
* and deleting or creating, drop it, else return name
* search for name in directory, to found or notfound
* notfound:
* if creating, return locked directory, leaving info on available slots
* else return error
* found:
* if at end of path and deleting, return information to allow delete
* if at end of path and rewriting (RENAME and LOCKPARENT), lock target
* inode and return info to allow rewrite
* if not at end, add name to cache; if at end and neither creating
* nor deleting, add name to cache
*/
int
ext2fs_lookup(void *v)
{
struct vop_lookup_v2_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
} */ *ap = v;
struct vnode *vdp = ap->a_dvp; /* vnode for directory being searched */
struct inode *dp = VTOI(vdp); /* inode for directory being searched */
struct buf *bp; /* a buffer of directory entries */
struct ext2fs_direct *ep; /* the current directory entry */
int entryoffsetinblock; /* offset of ep in bp's buffer */
enum {NONE, COMPACT, FOUND} slotstatus;
doff_t slotoffset; /* offset of area with free space */
int slotsize; /* size of area at slotoffset */
int slotfreespace; /* amount of space free in slot */
int slotneeded; /* size of the entry we're seeking */
int numdirpasses; /* strategy for directory search */
doff_t endsearch; /* offset to end directory search */
doff_t prevoff; /* prev entry dp->i_offset */
struct vnode *tdp; /* returned by vcache_get */
doff_t enduseful; /* pointer past last used dir slot */
u_long bmask; /* block offset mask */
int namlen, error;
struct vnode **vpp = ap->a_vpp;
struct componentname *cnp = ap->a_cnp;
kauth_cred_t cred = cnp->cn_cred;
int flags;
int nameiop = cnp->cn_nameiop;
struct ufsmount *ump = dp->i_ump;
int dirblksiz = ump->um_dirblksiz;
ino_t foundino;
struct ufs_lookup_results *results;
flags = cnp->cn_flags;
bp = NULL;
slotoffset = -1;
*vpp = NULL;
/*
* Produce the auxiliary lookup results into i_crap. Increment
* its serial number so elsewhere we can tell if we're using
* stale results. This should not be done this way. XXX.
*/
results = &dp->i_crap;
dp->i_crapcounter++;
/*
* Check accessiblity of directory.
*/
if ((error = VOP_ACCESS(vdp, VEXEC, cred)) != 0)
return (error);
if ((flags & ISLASTCN) && (vdp->v_mount->mnt_flag & MNT_RDONLY) &&
(cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
return (EROFS);
/*
* We now have a segment name to search for, and a directory to search.
*
* Before tediously performing a linear scan of the directory,
* check the name cache to see if the directory/name pair
* we are looking for is known already.
*/
if (cache_lookup(vdp, cnp->cn_nameptr, cnp->cn_namelen,
cnp->cn_nameiop, cnp->cn_flags, NULL, vpp)) {
return *vpp == NULLVP ? ENOENT : 0;
}
/*
* Suppress search for slots unless creating
* file and at end of pathname, in which case
* we watch for a place to put the new file in
* case it doesn't already exist.
*/
slotstatus = FOUND;
slotfreespace = slotsize = slotneeded = 0;
if ((nameiop == CREATE || nameiop == RENAME) &&
(flags & ISLASTCN)) {
slotstatus = NONE;
slotneeded = EXT2FS_DIRSIZ(cnp->cn_namelen);
}
/*
* If there is cached information on a previous search of
* this directory, pick up where we last left off.
* We cache only lookups as these are the most common
* and have the greatest payoff. Caching CREATE has little
* benefit as it usually must search the entire directory
* to determine that the entry does not exist. Caching the
* location of the last DELETE or RENAME has not reduced
* profiling time and hence has been removed in the interest
* of simplicity.
*/
bmask = vdp->v_mount->mnt_stat.f_iosize - 1;
if (nameiop != LOOKUP || results->ulr_diroff == 0 ||
results->ulr_diroff >= ext2fs_size(dp)) {
entryoffsetinblock = 0;
results->ulr_offset = 0;
numdirpasses = 1;
} else {
results->ulr_offset = results->ulr_diroff;
if ((entryoffsetinblock = results->ulr_offset & bmask) &&
(error = ext2fs_blkatoff(vdp, (off_t)results->ulr_offset, NULL, &bp)))
return (error);
numdirpasses = 2;
namecache_count_2passes();
}
prevoff = results->ulr_offset;
endsearch = roundup(ext2fs_size(dp), dirblksiz);
enduseful = 0;
searchloop:
while (results->ulr_offset < endsearch) {
if (curcpu()->ci_schedstate.spc_flags & SPCF_SHOULDYIELD)
preempt();
/*
* If necessary, get the next directory block.
*/
if ((results->ulr_offset & bmask) == 0) {
if (bp != NULL)
brelse(bp, 0);
error = ext2fs_blkatoff(vdp, (off_t)results->ulr_offset, NULL,
&bp);
if (error != 0)
return (error);
entryoffsetinblock = 0;
}
/*
* If still looking for a slot, and at a dirblksize
* boundary, have to start looking for free space again.
*/
if (slotstatus == NONE &&
(entryoffsetinblock & (dirblksiz - 1)) == 0) {
slotoffset = -1;
slotfreespace = 0;
}
/*
* Get pointer to next entry.
* Full validation checks are slow, so we only check
* enough to insure forward progress through the
* directory. Complete checks can be run by patching
* "dirchk" to be true.
*/
KASSERT(bp != NULL);
ep = (struct ext2fs_direct *)
((char *)bp->b_data + entryoffsetinblock);
if (ep->e2d_reclen == 0 ||
(dirchk &&
ext2fs_dirbadentry(vdp, ep, entryoffsetinblock))) {
int i;
ufs_dirbad(dp, results->ulr_offset, "mangled entry");
i = dirblksiz - (entryoffsetinblock & (dirblksiz - 1));
results->ulr_offset += i;
entryoffsetinblock += i;
continue;
}
/*
* If an appropriate sized slot has not yet been found,
* check to see if one is available. Also accumulate space
* in the current block so that we can determine if
* compaction is viable.
*/
if (slotstatus != FOUND) {
int size = fs2h16(ep->e2d_reclen);
if (ep->e2d_ino != 0)
size -= EXT2FS_DIRSIZ(ep->e2d_namlen);
if (size > 0) {
if (size >= slotneeded) {
slotstatus = FOUND;
slotoffset = results->ulr_offset;
slotsize = fs2h16(ep->e2d_reclen);
} else if (slotstatus == NONE) {
slotfreespace += size;
if (slotoffset == -1)
slotoffset = results->ulr_offset;
if (slotfreespace >= slotneeded) {
slotstatus = COMPACT;
slotsize = results->ulr_offset +
fs2h16(ep->e2d_reclen) -
slotoffset;
}
}
}
}
/*
* Check for a name match.
*/
if (ep->e2d_ino) {
namlen = ep->e2d_namlen;
if (namlen == cnp->cn_namelen &&
!memcmp(cnp->cn_nameptr, ep->e2d_name,
(unsigned)namlen)) {
/*
* Save directory entry's inode number and
* reclen in ndp->ni_ufs area, and release
* directory buffer.
*/
foundino = fs2h32(ep->e2d_ino);
results->ulr_reclen = fs2h16(ep->e2d_reclen);
goto found;
}
}
prevoff = results->ulr_offset;
results->ulr_offset += fs2h16(ep->e2d_reclen);
entryoffsetinblock += fs2h16(ep->e2d_reclen);
if (ep->e2d_ino)
enduseful = results->ulr_offset;
}
/* notfound: */
/*
* If we started in the middle of the directory and failed
* to find our target, we must check the beginning as well.
*/
if (numdirpasses == 2) {
numdirpasses--;
results->ulr_offset = 0;
endsearch = results->ulr_diroff;
goto searchloop;
}
if (bp != NULL)
brelse(bp, 0);
/*
* If creating, and at end of pathname and current
* directory has not been removed, then can consider
* allowing file to be created.
*/
if ((nameiop == CREATE || nameiop == RENAME) &&
(flags & ISLASTCN) && dp->i_e2fs_nlink != 0) {
/*
* Access for write is interpreted as allowing
* creation of files in the directory.
*/
error = VOP_ACCESS(vdp, VWRITE, cred);
if (error)
return (error);
/*
* Return an indication of where the new directory
* entry should be put. If we didn't find a slot,
* then set results->ulr_count to 0 indicating
* that the new slot belongs at the end of the
* directory. If we found a slot, then the new entry
* can be put in the range from results->ulr_offset to
* results->ulr_offset + results->ulr_count.
*/
if (slotstatus == NONE) {
results->ulr_offset = roundup(ext2fs_size(dp), dirblksiz);
results->ulr_count = 0;
enduseful = results->ulr_offset;
} else {
results->ulr_offset = slotoffset;
results->ulr_count = slotsize;
if (enduseful < slotoffset + slotsize)
enduseful = slotoffset + slotsize;
}
results->ulr_endoff = roundup(enduseful, dirblksiz);
#if 0
dp->i_flag |= IN_CHANGE | IN_UPDATE;
#endif
/*
* We return with the directory locked, so that
* the parameters we set up above will still be
* valid if we actually decide to do a direnter().
* We return ni_vp == NULL to indicate that the entry
* does not currently exist; we leave a pointer to
* the (locked) directory inode in ndp->ni_dvp.
*
* NB - if the directory is unlocked, then this
* information cannot be used.
*/
return (EJUSTRETURN);
}
/*
* Insert name into cache (as non-existent) if appropriate.
*/
if (nameiop != CREATE) {
cache_enter(vdp, *vpp, cnp->cn_nameptr, cnp->cn_namelen,
cnp->cn_flags);
}
return ENOENT;
found:
if (numdirpasses == 2)
namecache_count_pass2();
/*
* Check that directory length properly reflects presence
* of this entry.
*/
if (results->ulr_offset + EXT2FS_DIRSIZ(ep->e2d_namlen) > ext2fs_size(dp)) {
ufs_dirbad(dp, results->ulr_offset, "i_size too small");
error = ext2fs_setsize(dp,
results->ulr_offset + EXT2FS_DIRSIZ(ep->e2d_namlen));
if (error) {
brelse(bp, 0);
return (error);
}
dp->i_flag |= IN_CHANGE | IN_UPDATE;
uvm_vnp_setsize(vdp, ext2fs_size(dp));
}
brelse(bp, 0);
/*
* Found component in pathname.
* If the final component of path name, save information
* in the cache as to where the entry was found.
*/
if ((flags & ISLASTCN) && nameiop == LOOKUP)
results->ulr_diroff = results->ulr_offset &~ (dirblksiz - 1);
/*
* If deleting, and at end of pathname, return
* parameters which can be used to remove file.
* Lock the inode, being careful with ".".
*/
if (nameiop == DELETE && (flags & ISLASTCN)) {
/*
* Return pointer to current entry in results->ulr_offset,
* and distance past previous entry (if there
* is a previous entry in this block) in results->ulr_count.
* Save directory inode pointer in ndp->ni_dvp for dirremove().
*/
if ((results->ulr_offset & (dirblksiz - 1)) == 0)
results->ulr_count = 0;
else
results->ulr_count = results->ulr_offset - prevoff;
if (dp->i_number == foundino) {
vref(vdp);
tdp = vdp;
} else {
error = vcache_get(vdp->v_mount,
&foundino, sizeof(foundino), &tdp);
if (error)
return (error);
}
/*
* Write access to directory required to delete files.
*/
if ((error = VOP_ACCESS(vdp, VWRITE, cred)) != 0) {
vrele(tdp);
return (error);
}
/*
* If directory is "sticky", then user must own
* the directory, or the file in it, else she
* may not delete it (unless she's root). This
* implements append-only directories.
*/
if (dp->i_e2fs_mode & ISVTX) {
error = kauth_authorize_vnode(cred, KAUTH_VNODE_DELETE,
tdp, vdp, genfs_can_sticky(cred, dp->i_uid,
VTOI(tdp)->i_uid));
if (error) {
vrele(tdp);
return (EPERM);
}
}
*vpp = tdp;
return (0);
}
/*
* If rewriting (RENAME), return the inode and the
* information required to rewrite the present directory
* Must get inode of directory entry to verify it's a
* regular file, or empty directory.
*/
if (nameiop == RENAME && (flags & ISLASTCN)) {
error = VOP_ACCESS(vdp, VWRITE, cred);
if (error)
return (error);
/*
* Careful about locking second inode.
* This can only occur if the target is ".".
*/
if (dp->i_number == foundino)
return (EISDIR);
error = vcache_get(vdp->v_mount,
&foundino, sizeof(foundino), &tdp);
if (error)
return (error);
*vpp = tdp;
return (0);
}
if (dp->i_number == foundino) {
vref(vdp); /* we want ourself, ie "." */
*vpp = vdp;
} else {
error = vcache_get(vdp->v_mount,
&foundino, sizeof(foundino), &tdp);
if (error)
return (error);
*vpp = tdp;
}
/*
* Insert name into cache if appropriate.
*/
cache_enter(vdp, *vpp, cnp->cn_nameptr, cnp->cn_namelen, cnp->cn_flags);
return 0;
}
int
ext2fs_bmaparray(struct vnode *vp, int32_t bn, daddr_t *bnp,
struct indir *ap, int *nump, int *runp)
{
struct inode *ip;
struct buf *bp;
struct ufsmount *ump;
struct mount *mp;
struct vnode *devvp;
struct indir a[NIADDR+1], *xap;
int32_t daddr;
long metalbn;
int error, maxrun = 0, num;
ip = VTOI(vp);
mp = vp->v_mount;
ump = VFSTOUFS(mp);
#ifdef DIAGNOSTIC
if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL))
panic("ext2fs_bmaparray: invalid arguments");
#endif
if (runp) {
/*
* XXX
* If MAXBSIZE is the largest transfer the disks can handle,
* we probably want maxrun to be 1 block less so that we
* don't create a block larger than the device can handle.
*/
*runp = 0;
maxrun = MAXBSIZE / mp->mnt_stat.f_iosize - 1;
}
xap = ap == NULL ? a : ap;
if (!nump)
nump = #
if ((error = ufs_getlbns(vp, bn, xap, nump)) != 0)
return (error);
num = *nump;
if (num == 0) {
*bnp = blkptrtodb(ump, fs2h32(ip->i_e2fs_blocks[bn]));
if (*bnp == 0)
*bnp = -1;
else if (runp)
for (++bn; bn < NDADDR && *runp < maxrun &&
is_sequential(ump, fs2h32(ip->i_e2fs_blocks[bn - 1]),
fs2h32(ip->i_e2fs_blocks[bn]));
++bn, ++*runp);
return (0);
}
/* Get disk address out of indirect block array */
daddr = fs2h32(ip->i_e2fs_blocks[NDADDR + xap->in_off]);
devvp = VFSTOUFS(vp->v_mount)->um_devvp;
#ifdef DIAGNOSTIC
if (num > NIADDR + 1 || num < 1) {
printf("ext2fs_bmaparray: num=%d\n", num);
panic("ext2fs_bmaparray: num");
}
#endif
for (bp = NULL, ++xap; --num; ++xap) {
/*
* Exit the loop if there is no disk address assigned yet and
* the indirect block isn't in the cache, or if we were
* looking for an indirect block and we've found it.
*/
metalbn = xap->in_lbn;
if ((daddr == 0 && !incore(vp, metalbn)) || metalbn == bn)
break;
/*
* If we get here, we've either got the block in the cache
* or we have a disk address for it, go fetch it.
*/
if (bp)
brelse(bp);
xap->in_exists = 1;
bp = getblk(vp, metalbn, mp->mnt_stat.f_iosize, 0, 0);
if (bp->b_flags & (B_DONE | B_DELWRI)) {
;
}
#ifdef DIAGNOSTIC
else if (!daddr)
panic("ext2fs_bmaparry: indirect block not in cache");
#endif
else {
bp->b_blkno = blkptrtodb(ump, daddr);
bp->b_flags |= B_READ;
VOP_STRATEGY(bp);
curproc->p_ru.ru_inblock++; /* XXX */
bcstats.pendingreads++;
if ((error = biowait(bp)) != 0) {
brelse(bp);
return (error);
}
}
daddr = fs2h32(((int32_t *)bp->b_data)[xap->in_off]);
if (num == 1 && daddr && runp)
for (bn = xap->in_off + 1;
bn < MNINDIR(ump) && *runp < maxrun &&
is_sequential(ump, ((int32_t *)bp->b_data)[bn - 1],
((int32_t *)bp->b_data)[bn]);
++bn, ++*runp);
}
if (bp)
brelse(bp);
daddr = blkptrtodb(ump, daddr);
*bnp = daddr == 0 ? -1 : daddr;
return (0);
}
static int
pass2check(struct inodesc *idesc)
{
struct ext2fs_direct *dirp = idesc->id_dirp;
struct inoinfo *inp;
int n, entrysize, ret = 0;
struct ext2fs_dinode *dp;
const char *errmsg;
struct ext2fs_direct proto;
char namebuf[MAXPATHLEN + 1];
char pathbuf[MAXPATHLEN + 1];
/*
* check for "."
*/
if (idesc->id_entryno != 0)
goto chk1;
if (fs2h32(dirp->e2d_ino) != 0 && dirp->e2d_namlen == 1 &&
dirp->e2d_name[0] == '.') {
if (fs2h32(dirp->e2d_ino) != idesc->id_number) {
direrror(idesc->id_number, "BAD INODE NUMBER FOR '.'");
dirp->e2d_ino = h2fs32(idesc->id_number);
if (reply("FIX") == 1)
ret |= ALTERED;
}
if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
(sblock.e2fs.e2fs_features_incompat & EXT2F_INCOMPAT_FTYPE)
&& (dirp->e2d_type != EXT2_FT_DIR)) {
direrror(idesc->id_number, "BAD TYPE VALUE FOR '.'");
dirp->e2d_type = EXT2_FT_DIR;
if (reply("FIX") == 1)
ret |= ALTERED;
}
goto chk1;
}
direrror(idesc->id_number, "MISSING '.'");
proto.e2d_ino = h2fs32(idesc->id_number);
proto.e2d_namlen = 1;
if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
(sblock.e2fs.e2fs_features_incompat & EXT2F_INCOMPAT_FTYPE))
proto.e2d_type = EXT2_FT_DIR;
else
proto.e2d_type = 0;
(void)strlcpy(proto.e2d_name, ".", sizeof(proto.e2d_name));
entrysize = EXT2FS_DIRSIZ(proto.e2d_namlen);
if (fs2h32(dirp->e2d_ino) != 0 && strcmp(dirp->e2d_name, "..") != 0) {
pfatal("CANNOT FIX, FIRST ENTRY IN DIRECTORY CONTAINS %s\n",
dirp->e2d_name);
} else if (fs2h16(dirp->e2d_reclen) < entrysize) {
pfatal("CANNOT FIX, INSUFFICIENT SPACE TO ADD '.'\n");
} else if (fs2h16(dirp->e2d_reclen) < 2 * entrysize) {
proto.e2d_reclen = dirp->e2d_reclen;
memcpy(dirp, &proto, (size_t)entrysize);
if (reply("FIX") == 1)
ret |= ALTERED;
} else {
n = fs2h16(dirp->e2d_reclen) - entrysize;
proto.e2d_reclen = h2fs16(entrysize);
memcpy(dirp, &proto, (size_t)entrysize);
idesc->id_entryno++;
lncntp[fs2h32(dirp->e2d_ino)]--;
dirp = (struct ext2fs_direct *)((char *)(dirp) + entrysize);
memset(dirp, 0, (size_t)n);
dirp->e2d_reclen = h2fs16(n);
if (reply("FIX") == 1)
ret |= ALTERED;
}
chk1:
if (idesc->id_entryno > 1)
goto chk2;
inp = getinoinfo(idesc->id_number);
proto.e2d_ino = h2fs32(inp->i_parent);
proto.e2d_namlen = 2;
if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
(sblock.e2fs.e2fs_features_incompat & EXT2F_INCOMPAT_FTYPE))
proto.e2d_type = EXT2_FT_DIR;
else
proto.e2d_type = 0;
(void)strlcpy(proto.e2d_name, "..", sizeof(proto.e2d_name));
entrysize = EXT2FS_DIRSIZ(2);
if (idesc->id_entryno == 0) {
n = EXT2FS_DIRSIZ(dirp->e2d_namlen);
if (fs2h16(dirp->e2d_reclen) < n + entrysize)
goto chk2;
proto.e2d_reclen = h2fs16(fs2h16(dirp->e2d_reclen) - n);
dirp->e2d_reclen = h2fs16(n);
idesc->id_entryno++;
lncntp[fs2h32(dirp->e2d_ino)]--;
dirp = (struct ext2fs_direct *)((char *)(dirp) + n);
memset(dirp, 0, (size_t)fs2h16(proto.e2d_reclen));
dirp->e2d_reclen = proto.e2d_reclen;
}
if (fs2h32(dirp->e2d_ino) != 0 &&
dirp->e2d_namlen == 2 &&
strncmp(dirp->e2d_name, "..", 2) == 0) {
inp->i_dotdot = fs2h32(dirp->e2d_ino);
if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
(sblock.e2fs.e2fs_features_incompat & EXT2F_INCOMPAT_FTYPE)
&& dirp->e2d_type != EXT2_FT_DIR) {
direrror(idesc->id_number, "BAD TYPE VALUE FOR '..'");
dirp->e2d_type = EXT2_FT_DIR;
if (reply("FIX") == 1)
ret |= ALTERED;
}
goto chk2;
}
if (fs2h32(dirp->e2d_ino) != 0 &&
dirp->e2d_namlen == 1 &&
strncmp(dirp->e2d_name, ".", 1) != 0) {
fileerror(inp->i_parent, idesc->id_number, "MISSING '..'");
pfatal("CANNOT FIX, SECOND ENTRY IN DIRECTORY CONTAINS %s\n",
dirp->e2d_name);
inp->i_dotdot = (ino_t)-1;
} else if (fs2h16(dirp->e2d_reclen) < entrysize) {
fileerror(inp->i_parent, idesc->id_number, "MISSING '..'");
pfatal("CANNOT FIX, INSUFFICIENT SPACE TO ADD '..'\n");
inp->i_dotdot = (ino_t)-1;
} else if (inp->i_parent != 0) {
/*
* We know the parent, so fix now.
*/
inp->i_dotdot = inp->i_parent;
fileerror(inp->i_parent, idesc->id_number, "MISSING '..'");
proto.e2d_reclen = dirp->e2d_reclen;
memcpy(dirp, &proto, (size_t)entrysize);
if (reply("FIX") == 1)
ret |= ALTERED;
}
idesc->id_entryno++;
if (fs2h32(dirp->e2d_ino) != 0)
lncntp[fs2h32(dirp->e2d_ino)]--;
return (ret|KEEPON);
chk2:
if (fs2h32(dirp->e2d_ino) == 0)
return (ret|KEEPON);
if (dirp->e2d_namlen <= 2 &&
dirp->e2d_name[0] == '.' &&
idesc->id_entryno >= 2) {
if (dirp->e2d_namlen == 1) {
direrror(idesc->id_number, "EXTRA '.' ENTRY");
dirp->e2d_ino = 0;
if (reply("FIX") == 1)
ret |= ALTERED;
return (KEEPON | ret);
}
if (dirp->e2d_name[1] == '.') {
direrror(idesc->id_number, "EXTRA '..' ENTRY");
dirp->e2d_ino = 0;
if (reply("FIX") == 1)
ret |= ALTERED;
return (KEEPON | ret);
}
}
idesc->id_entryno++;
n = 0;
if (fs2h32(dirp->e2d_ino) > maxino ||
(fs2h32(dirp->e2d_ino) < EXT2_FIRSTINO &&
fs2h32(dirp->e2d_ino) != EXT2_ROOTINO)) {
fileerror(idesc->id_number, fs2h32(dirp->e2d_ino), "I OUT OF RANGE");
n = reply("REMOVE");
} else {
again:
switch (statemap[fs2h32(dirp->e2d_ino)]) {
case USTATE:
if (idesc->id_entryno <= 2)
break;
fileerror(idesc->id_number, fs2h32(dirp->e2d_ino), "UNALLOCATED");
n = reply("REMOVE");
break;
case DCLEAR:
case FCLEAR:
if (idesc->id_entryno <= 2)
break;
if (statemap[fs2h32(dirp->e2d_ino)] == FCLEAR)
errmsg = "DUP/BAD";
else if (!preen)
errmsg = "ZERO LENGTH DIRECTORY";
else {
n = 1;
break;
}
fileerror(idesc->id_number, fs2h32(dirp->e2d_ino), errmsg);
if ((n = reply("REMOVE")) == 1)
break;
dp = ginode(fs2h32(dirp->e2d_ino));
statemap[fs2h32(dirp->e2d_ino)] =
(fs2h16(dp->e2di_mode) & IFMT) == IFDIR ? DSTATE : FSTATE;
lncntp[fs2h32(dirp->e2d_ino)] = fs2h16(dp->e2di_nlink);
goto again;
case DSTATE:
case DFOUND:
inp = getinoinfo(fs2h32(dirp->e2d_ino));
if (inp->i_parent != 0 && idesc->id_entryno > 2) {
getpathname(pathbuf, sizeof(pathbuf),
idesc->id_number, idesc->id_number);
getpathname(namebuf, sizeof(namebuf),
fs2h32(dirp->e2d_ino),
fs2h32(dirp->e2d_ino));
pwarn("%s %s %s\n", pathbuf,
"IS AN EXTRANEOUS HARD LINK TO DIRECTORY",
namebuf);
if (preen)
printf(" (IGNORED)\n");
else if ((n = reply("REMOVE")) == 1)
break;
}
if (idesc->id_entryno > 2)
inp->i_parent = idesc->id_number;
/* fall through */
case FSTATE:
if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
(sblock.e2fs.e2fs_features_incompat &
EXT2F_INCOMPAT_FTYPE) &&
dirp->e2d_type !=
inot2ext2dt(typemap[fs2h32(dirp->e2d_ino)])) {
dirp->e2d_type =
inot2ext2dt(typemap[fs2h32(dirp->e2d_ino)]);
fileerror(idesc->id_number,
fs2h32(dirp->e2d_ino),
"BAD TYPE VALUE");
if (reply("FIX") == 1)
ret |= ALTERED;
}
lncntp[fs2h32(dirp->e2d_ino)]--;
break;
default:
errexit("BAD STATE %d FOR INODE I=%d",
statemap[fs2h32(dirp->e2d_ino)], fs2h32(dirp->e2d_ino));
}
}
if (n == 0)
return (ret|KEEPON);
dirp->e2d_ino = 0;
return (ret|KEEPON|ALTERED);
}
/*
* Given an offset in a file, find the disk block number that
* contains that block.
*/
static int
block_map(struct open_file *f, indp_t file_block, indp_t *disk_block_p)
{
struct file *fp = (struct file *)f->f_fsdata;
struct m_ext2fs *fs = fp->f_fs;
uint level;
indp_t ind_cache;
indp_t ind_block_num;
size_t rsize;
int rc;
indp_t *buf = (void *)fp->f_buf;
/*
* Index structure of an inode:
*
* e2di_blocks[0..EXT2FS_NDADDR-1]
* hold block numbers for blocks
* 0..EXT2FS_NDADDR-1
*
* e2di_blocks[EXT2FS_NDADDR+0]
* block EXT2FS_NDADDR+0 is the single indirect block
* holds block numbers for blocks
* EXT2FS_NDADDR .. EXT2FS_NDADDR + EXT2_NINDIR(fs)-1
*
* e2di_blocks[EXT2FS_NDADDR+1]
* block EXT2FS_NDADDR+1 is the double indirect block
* holds block numbers for INDEX blocks for blocks
* EXT2FS_NDADDR + EXT2_NINDIR(fs) ..
* EXT2FS_NDADDR + EXT2_NINDIR(fs) + EXT2_NINDIR(fs)**2 - 1
*
* e2di_blocks[EXT2FS_NDADDR+2]
* block EXT2FS_NDADDR+2 is the triple indirect block
* holds block numbers for double-indirect
* blocks for blocks
* EXT2FS_NDADDR + EXT2_NINDIR(fs) + EXT2_NINDIR(fs)**2 ..
* EXT2FS_NDADDR + EXT2_NINDIR(fs) + EXT2_NINDIR(fs)**2
* + EXT2_NINDIR(fs)**3 - 1
*/
if (file_block < EXT2FS_NDADDR) {
/* Direct block. */
*disk_block_p = fs2h32(fp->f_di.e2di_blocks[file_block]);
return 0;
}
file_block -= EXT2FS_NDADDR;
ind_cache = file_block >> LN2_IND_CACHE_SZ;
if (ind_cache == fp->f_ind_cache_block) {
*disk_block_p =
fs2h32(fp->f_ind_cache[file_block & IND_CACHE_MASK]);
return 0;
}
for (level = 0;;) {
level += fp->f_nishift;
if (file_block < (indp_t)1 << level)
break;
if (level > EXT2FS_NIADDR * fp->f_nishift)
/* Block number too high */
return EFBIG;
file_block -= (indp_t)1 << level;
}
ind_block_num =
fs2h32(fp->f_di.e2di_blocks[EXT2FS_NDADDR +
(level / fp->f_nishift - 1)]);
for (;;) {
level -= fp->f_nishift;
if (ind_block_num == 0) {
*disk_block_p = 0; /* missing */
return 0;
}
twiddle();
/*
* If we were feeling brave, we could work out the number
* of the disk sector and read a single disk sector instead
* of a filesystem block.
* However we don't do this very often anyway...
*/
rc = DEV_STRATEGY(f->f_dev)(f->f_devdata, F_READ,
FSBTODB(fp->f_fs, ind_block_num), fs->e2fs_bsize,
buf, &rsize);
if (rc)
return rc;
if (rsize != fs->e2fs_bsize)
return EIO;
ind_block_num = fs2h32(buf[file_block >> level]);
if (level == 0)
break;
file_block &= (1 << level) - 1;
}
/* Save the part of the block that contains this sector */
memcpy(fp->f_ind_cache, &buf[file_block & ~IND_CACHE_MASK],
IND_CACHE_SZ * sizeof fp->f_ind_cache[0]);
fp->f_ind_cache_block = ind_cache;
*disk_block_p = ind_block_num;
return 0;
}
__compactcall void
ext2fs_ls(struct open_file *f, const char *pattern)
{
struct file *fp = (struct file *)f->f_fsdata;
size_t block_size = fp->f_fs->e2fs_bsize;
char *buf;
size_t buf_size;
entry_t *names = 0, *n, **np;
fp->f_seekp = 0;
while (fp->f_seekp < (off_t)fp->f_di.e2di_size) {
struct ext2fs_direct *dp, *edp;
int rc = buf_read_file(f, &buf, &buf_size);
if (rc)
goto out;
if (buf_size != block_size || buf_size == 0)
goto out;
dp = (struct ext2fs_direct *)buf;
edp = (struct ext2fs_direct *)(buf + buf_size);
for (; dp < edp;
dp = (void *)((char *)dp + fs2h16(dp->e2d_reclen))) {
const char *t;
if (fs2h16(dp->e2d_reclen) <= 0)
goto out;
if (fs2h32(dp->e2d_ino) == 0)
continue;
if (dp->e2d_type >= NELEM(typestr) ||
!(t = typestr[dp->e2d_type])) {
/*
* This does not handle "old"
* filesystems properly. On little
* endian machines, we get a bogus
* type name if the namlen matches a
* valid type identifier. We could
* check if we read namlen "0" and
* handle this case specially, if
* there were a pressing need...
*/
printf("bad dir entry\n");
goto out;
}
if (pattern && !fnmatch(dp->e2d_name, pattern))
continue;
n = alloc(sizeof *n + strlen(dp->e2d_name));
if (!n) {
printf("%d: %s (%s)\n",
fs2h32(dp->e2d_ino), dp->e2d_name, t);
continue;
}
n->e_ino = fs2h32(dp->e2d_ino);
n->e_type = dp->e2d_type;
strcpy(n->e_name, dp->e2d_name);
for (np = &names; *np; np = &(*np)->e_next) {
if (strcmp(n->e_name, (*np)->e_name) < 0)
break;
}
n->e_next = *np;
*np = n;
}
fp->f_seekp += buf_size;
}
if (names) {
entry_t *p_names = names;
do {
n = p_names;
printf("%d: %s (%s)\n",
n->e_ino, n->e_name, typestr[n->e_type]);
p_names = n->e_next;
} while (p_names);
} else {
printf("not found\n");
}
out:
if (names) {
do {
n = names;
names = n->e_next;
dealloc(n, 0);
} while (names);
}
return;
}
void
pass5(void)
{
int c;
struct m_ext2fs *fs = &sblock;
daddr_t dbase, dmax;
daddr_t d;
long i, j;
struct inodesc idesc[3];
struct bufarea *ino_bitmap = NULL, *blk_bitmap = NULL;
char *ibmap, *bbmap;
u_int32_t cs_ndir, cs_nbfree, cs_nifree;
char msg[255];
cs_ndir = 0;
cs_nbfree = 0;
cs_nifree = 0;
ibmap = malloc(fs->e2fs_bsize);
bbmap = malloc(fs->e2fs_bsize);
if (ibmap == NULL || bbmap == NULL) {
errexit("out of memory\n");
}
for (c = 0; c < fs->e2fs_ncg; c++) {
u_int32_t nbfree = 0;
u_int32_t nifree = 0;
u_int32_t ndirs = 0;
nbfree = 0;
nifree = fs->e2fs.e2fs_ipg;
ndirs = 0;
if (blk_bitmap == NULL) {
blk_bitmap = getdatablk(fs2h32(fs->e2fs_gd[c].ext2bgd_b_bitmap),
fs->e2fs_bsize);
} else {
getblk(blk_bitmap, fs2h32(fs->e2fs_gd[c].ext2bgd_b_bitmap),
fs->e2fs_bsize);
}
if (ino_bitmap == NULL) {
ino_bitmap = getdatablk(fs2h32(fs->e2fs_gd[c].ext2bgd_i_bitmap),
fs->e2fs_bsize);
} else {
getblk(ino_bitmap, fs2h32(fs->e2fs_gd[c].ext2bgd_i_bitmap),
fs->e2fs_bsize);
}
memset(bbmap, 0, fs->e2fs_bsize);
memset(ibmap, 0, fs->e2fs_bsize);
memset(&idesc[0], 0, sizeof idesc);
for (i = 0; i < 3; i++) {
idesc[i].id_type = ADDR;
}
j = fs->e2fs.e2fs_ipg * c + 1;
for (i = 0; i < fs->e2fs.e2fs_ipg; j++, i++) {
if ((j < EXT2_FIRSTINO) && (j != EXT2_ROOTINO)) {
setbit(ibmap, i);
nifree--;
continue;
}
if (j > fs->e2fs.e2fs_icount) {
setbit(ibmap, i);
continue;
}
switch (statemap[j]) {
case USTATE:
break;
case DSTATE:
case DCLEAR:
case DFOUND:
ndirs++;
/* fall through */
case FSTATE:
case FCLEAR:
nifree--;
setbit(ibmap, i);
break;
default:
errexit("BAD STATE %d FOR INODE I=%ld\n",
statemap[j], j);
}
}
/* fill in unused par of the inode map */
for (i = fs->e2fs.e2fs_ipg / NBBY; i < fs->e2fs_bsize; i++)
ibmap[i] = 0xff;
dbase = c * sblock.e2fs.e2fs_bpg +
sblock.e2fs.e2fs_first_dblock;
dmax = (c+1) * sblock.e2fs.e2fs_bpg +
sblock.e2fs.e2fs_first_dblock;
for (i = 0, d = dbase;
d < dmax;
d ++, i ++) {
if (testbmap(d) || d >= sblock.e2fs.e2fs_bcount) {
setbit(bbmap, i);
continue;
} else {
nbfree++;
}
}
cs_nbfree += nbfree;
cs_nifree += nifree;
cs_ndir += ndirs;
if (debug && (fs2h16(fs->e2fs_gd[c].ext2bgd_nbfree) != nbfree ||
fs2h16(fs->e2fs_gd[c].ext2bgd_nifree) != nifree ||
fs2h16(fs->e2fs_gd[c].ext2bgd_ndirs) != ndirs)) {
printf("summary info for cg %d is %d, %d, %d,"
"should be %d, %d, %d\n", c,
fs2h16(fs->e2fs_gd[c].ext2bgd_nbfree),
fs2h16(fs->e2fs_gd[c].ext2bgd_nifree),
fs2h16(fs->e2fs_gd[c].ext2bgd_ndirs),
nbfree,
nifree,
ndirs);
}
(void)snprintf(msg, sizeof(msg),
"SUMMARY INFORMATIONS WRONG FOR CG #%d", c);
if ((fs2h16(fs->e2fs_gd[c].ext2bgd_nbfree) != nbfree ||
fs2h16(fs->e2fs_gd[c].ext2bgd_nifree) != nifree ||
fs2h16(fs->e2fs_gd[c].ext2bgd_ndirs) != ndirs) &&
dofix(&idesc[0], msg)) {
fs->e2fs_gd[c].ext2bgd_nbfree = h2fs16(nbfree);
fs->e2fs_gd[c].ext2bgd_nifree = h2fs16(nifree);
fs->e2fs_gd[c].ext2bgd_ndirs = h2fs16(ndirs);
sbdirty();
}
if (debug && memcmp(blk_bitmap->b_un.b_buf, bbmap, fs->e2fs_bsize)) {
printf("blk_bitmap:\n");
print_bmap(blk_bitmap->b_un.b_buf, fs->e2fs_bsize);
printf("bbmap:\n");
print_bmap(bbmap, fs->e2fs_bsize);
}
(void)snprintf(msg, sizeof(msg),
"BLK(S) MISSING IN BIT MAPS #%d", c);
if (memcmp(blk_bitmap->b_un.b_buf, bbmap, fs->e2fs_bsize) &&
dofix(&idesc[1], msg)) {
memcpy(blk_bitmap->b_un.b_buf, bbmap, fs->e2fs_bsize);
dirty(blk_bitmap);
}
if (debug && memcmp(ino_bitmap->b_un.b_buf, ibmap, fs->e2fs_bsize)) {
printf("ino_bitmap:\n");
print_bmap(ino_bitmap->b_un.b_buf, fs->e2fs_bsize);
printf("ibmap:\n");
print_bmap(ibmap, fs->e2fs_bsize);
}
(void)snprintf(msg, sizeof(msg),
"INODE(S) MISSING IN BIT MAPS #%d", c);
if (memcmp(ino_bitmap->b_un.b_buf, ibmap, fs->e2fs_bsize) &&
dofix(&idesc[1], msg)) {
memcpy(ino_bitmap->b_un.b_buf, ibmap, fs->e2fs_bsize);
dirty(ino_bitmap);
}
}
if (debug && (fs->e2fs.e2fs_fbcount != cs_nbfree ||
fs->e2fs.e2fs_ficount != cs_nifree)) {
printf("summary info bad in superblock: %d, %d should be %d, %d\n",
fs->e2fs.e2fs_fbcount, fs->e2fs.e2fs_ficount,
cs_nbfree, cs_nifree);
}
if ((fs->e2fs.e2fs_fbcount != cs_nbfree ||
fs->e2fs.e2fs_ficount != cs_nifree)
&& dofix(&idesc[0], "SUPERBLK SUMMARY INFORMATION BAD")) {
fs->e2fs.e2fs_fbcount = cs_nbfree;
fs->e2fs.e2fs_ficount = cs_nifree;
sbdirty();
}
free(ibmap);
free(bbmap);
}
/*
* Given an offset in a file, find the disk block number (not zone!)
* that contains that block.
*/
static int
block_map(struct open_file *f, block_t file_block, block_t *disk_block_p)
{
struct file *fp = (struct file *)f->f_fsdata;
struct mfs_sblock *fs = fp->f_fs;
uint level;
block_t ind_cache;
block_t ind_block_num;
zone_t zone;
size_t rsize;
int rc;
int boff;
int scale = fs->mfs_log_zone_size; /* for block-zone conversion */
block_t *buf = (void *)fp->f_buf;
/*
* Index structure of an inode:
*
* mdi_blocks[0..NR_DZONES-1]
* hold zone numbers for zones
* 0..NR_DZONES-1
*
* mdi_blocks[NR_DZONES+0]
* block NDADDR+0 is the single indirect block
* holds zone numbers for zones
* NR_DZONES .. NR_DZONES + MFS_NINDIR(fs)-1
*
* mdi_blocks[NR_DZONES+1]
* block NDADDR+1 is the double indirect block
* holds zone numbers for INDEX blocks for zones
* NR_DZONES + MFS_NINDIR(fs) ..
* NR_TZONES + MFS_NINDIR(fs) + MFS_NINDIR(fs)**2 - 1
*/
zone = file_block >> scale;
boff = (int) (file_block - (zone << scale) ); /* relative blk in zone */
if (zone < NR_DZONES) {
/* Direct zone */
zone_t z = fs2h32(fp->f_di.mdi_zone[zone]);
if (z == NO_ZONE) {
*disk_block_p = NO_BLOCK;
return 0;
}
*disk_block_p = (block_t) ((z << scale) + boff);
return 0;
}
zone -= NR_DZONES;
ind_cache = zone >> LN2_IND_CACHE_SZ;
if (ind_cache == fp->f_ind_cache_block) {
*disk_block_p =
fs2h32(fp->f_ind_cache[zone & IND_CACHE_MASK]);
return 0;
}
for (level = 0;;) {
level += fp->f_nishift;
if (zone < (block_t)1 << level)
break;
if (level > NIADDR * fp->f_nishift)
/* Zone number too high */
return EFBIG;
zone -= (block_t)1 << level;
}
ind_block_num =
fs2h32(fp->f_di.mdi_zone[NR_DZONES + (level / fp->f_nishift - 1)]);
for (;;) {
level -= fp->f_nishift;
if (ind_block_num == 0) {
*disk_block_p = NO_BLOCK; /* missing */
return 0;
}
twiddle();
/*
* If we were feeling brave, we could work out the number
* of the disk sector and read a single disk sector instead
* of a filesystem block.
* However we don't do this very often anyway...
*/
rc = DEV_STRATEGY(f->f_dev)(f->f_devdata, F_READ,
FSBTODB(fs, ind_block_num), fs->mfs_block_size,
buf, &rsize);
if (rc)
return rc;
if (rsize != fs->mfs_block_size)
return EIO;
ind_block_num = fs2h32(buf[zone >> level]);
if (level == 0)
break;
zone &= (1 << level) - 1;
}
/* Save the part of the block that contains this sector */
memcpy(fp->f_ind_cache, &buf[zone & ~IND_CACHE_MASK],
IND_CACHE_SZ * sizeof fp->f_ind_cache[0]);
fp->f_ind_cache_block = ind_cache;
zone = (zone_t)ind_block_num;
*disk_block_p = (block_t)((zone << scale) + boff);
return 0;
}
/*
* Check if source directory is in the path of the target directory.
* Target is supplied locked, source is unlocked.
* The target is always vput before returning.
*/
int
ext2fs_checkpath(struct inode *source, struct inode *target,
struct ucred *cred)
{
struct vnode *vp;
int error, rootino, namlen;
struct ext2fs_dirtemplate dirbuf;
u_int32_t ino;
vp = ITOV(target);
if (target->i_number == source->i_number) {
error = EEXIST;
goto out;
}
rootino = ROOTINO;
error = 0;
if (target->i_number == rootino)
goto out;
for (;;) {
if (vp->v_type != VDIR) {
error = ENOTDIR;
break;
}
error = vn_rdwr(UIO_READ, vp, (caddr_t)&dirbuf,
sizeof (struct ext2fs_dirtemplate), (off_t)0,
UIO_SYSSPACE, IO_NODELOCKED, cred, NULL,
curproc);
if (error != 0)
break;
namlen = dirbuf.dotdot_namlen;
if (namlen != 2 ||
dirbuf.dotdot_name[0] != '.' ||
dirbuf.dotdot_name[1] != '.') {
error = ENOTDIR;
break;
}
ino = fs2h32(dirbuf.dotdot_ino);
if (ino == source->i_number) {
error = EINVAL;
break;
}
if (ino == rootino)
break;
vput(vp);
error = VFS_VGET(vp->v_mount, ino, &vp);
if (error != 0) {
vp = NULL;
break;
}
}
out:
if (error == ENOTDIR) {
printf("checkpath: .. not a directory\n");
panic("checkpath");
}
if (vp != NULL)
vput(vp);
return (error);
}
/*
* Read in the super block and its summary info, convert to host byte order.
*/
static int
readsb(int listerr)
{
daddr_t super = bflag ? bflag : SBOFF / dev_bsize;
if (bread(fsreadfd, (char *)sblk.b_un.b_fs, super, (long)SBSIZE) != 0)
return 0;
sblk.b_bno = super;
sblk.b_size = SBSIZE;
/* Copy the superblock in memory */
e2fs_sbload(sblk.b_un.b_fs, &sblock.e2fs);
/*
* run a few consistency checks of the super block
*/
if (sblock.e2fs.e2fs_magic != E2FS_MAGIC) {
badsb(listerr, "MAGIC NUMBER WRONG");
return 0;
}
if (sblock.e2fs.e2fs_log_bsize > 2) {
badsb(listerr, "BAD LOG_BSIZE");
return 0;
}
if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
(!powerof2(sblock.e2fs.e2fs_inode_size) ||
sblock.e2fs.e2fs_inode_size < EXT2_REV0_DINODE_SIZE ||
sblock.e2fs.e2fs_inode_size >
(1024 << sblock.e2fs.e2fs_log_bsize))) {
badsb(listerr, "BAD INODE_SIZE");
return 0;
}
/* compute the dynamic fields of the in-memory sb */
/* compute dynamic sb infos */
sblock.e2fs_ncg =
howmany(sblock.e2fs.e2fs_bcount - sblock.e2fs.e2fs_first_dblock,
sblock.e2fs.e2fs_bpg);
/* XXX assume hw bsize = 512 */
sblock.e2fs_fsbtodb = sblock.e2fs.e2fs_log_bsize + 1;
sblock.e2fs_bsize = 1024 << sblock.e2fs.e2fs_log_bsize;
sblock.e2fs_bshift = LOG_MINBSIZE + sblock.e2fs.e2fs_log_bsize;
sblock.e2fs_qbmask = sblock.e2fs_bsize - 1;
sblock.e2fs_bmask = ~sblock.e2fs_qbmask;
sblock.e2fs_ngdb = howmany(sblock.e2fs_ncg,
sblock.e2fs_bsize / sizeof(struct ext2_gd));
sblock.e2fs_ipb = sblock.e2fs_bsize / EXT2_DINODE_SIZE(&sblock);
sblock.e2fs_itpg = howmany(sblock.e2fs.e2fs_ipg, sblock.e2fs_ipb);
/*
* Compute block size that the filesystem is based on,
* according to fsbtodb, and adjust superblock block number
* so we can tell if this is an alternate later.
*/
super *= dev_bsize;
dev_bsize = sblock.e2fs_bsize / EXT2_FSBTODB(&sblock, 1);
sblk.b_bno = super / dev_bsize;
if (sblock.e2fs_ncg == 1) {
/* no alternate superblock; assume it's okay */
havesb = 1;
return 1;
}
getblk(&asblk, 1 * sblock.e2fs.e2fs_bpg + sblock.e2fs.e2fs_first_dblock,
(long)SBSIZE);
if (asblk.b_errs)
return 0;
if (bflag) {
havesb = 1;
return 1;
}
/*
* Set all possible fields that could differ, then do check
* of whole super block against an alternate super block.
* When an alternate super-block is specified this check is skipped.
*/
asblk.b_un.b_fs->e2fs_rbcount = sblk.b_un.b_fs->e2fs_rbcount;
asblk.b_un.b_fs->e2fs_fbcount = sblk.b_un.b_fs->e2fs_fbcount;
asblk.b_un.b_fs->e2fs_ficount = sblk.b_un.b_fs->e2fs_ficount;
asblk.b_un.b_fs->e2fs_mtime = sblk.b_un.b_fs->e2fs_mtime;
asblk.b_un.b_fs->e2fs_wtime = sblk.b_un.b_fs->e2fs_wtime;
asblk.b_un.b_fs->e2fs_mnt_count = sblk.b_un.b_fs->e2fs_mnt_count;
asblk.b_un.b_fs->e2fs_max_mnt_count =
sblk.b_un.b_fs->e2fs_max_mnt_count;
asblk.b_un.b_fs->e2fs_state = sblk.b_un.b_fs->e2fs_state;
asblk.b_un.b_fs->e2fs_beh = sblk.b_un.b_fs->e2fs_beh;
asblk.b_un.b_fs->e2fs_lastfsck = sblk.b_un.b_fs->e2fs_lastfsck;
asblk.b_un.b_fs->e2fs_fsckintv = sblk.b_un.b_fs->e2fs_fsckintv;
asblk.b_un.b_fs->e2fs_ruid = sblk.b_un.b_fs->e2fs_ruid;
asblk.b_un.b_fs->e2fs_rgid = sblk.b_un.b_fs->e2fs_rgid;
asblk.b_un.b_fs->e2fs_block_group_nr =
sblk.b_un.b_fs->e2fs_block_group_nr;
asblk.b_un.b_fs->e2fs_features_rocompat &= ~EXT2F_ROCOMPAT_LARGEFILE;
asblk.b_un.b_fs->e2fs_features_rocompat |=
sblk.b_un.b_fs->e2fs_features_rocompat & EXT2F_ROCOMPAT_LARGEFILE;
if (sblock.e2fs.e2fs_rev > E2FS_REV0 &&
((sblock.e2fs.e2fs_features_incompat & ~EXT2F_INCOMPAT_SUPP_FSCK) ||
(sblock.e2fs.e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP_FSCK))) {
if (debug) {
printf("compat 0x%08x, incompat 0x%08x, compat_ro "
"0x%08x\n",
sblock.e2fs.e2fs_features_compat,
sblock.e2fs.e2fs_features_incompat,
sblock.e2fs.e2fs_features_rocompat);
if ((sblock.e2fs.e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP_FSCK)) {
char buf[512];
snprintb(buf, sizeof(buf), EXT2F_ROCOMPAT_BITS,
sblock.e2fs.e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP_FSCK);
printf("unsupported rocompat features: %s\n", buf);
}
if ((sblock.e2fs.e2fs_features_incompat & ~EXT2F_INCOMPAT_SUPP_FSCK)) {
char buf[512];
snprintb(buf, sizeof(buf), EXT2F_INCOMPAT_BITS,
sblock.e2fs.e2fs_features_incompat & ~EXT2F_INCOMPAT_SUPP_FSCK);
printf("unsupported incompat features: %s\n", buf);
}
}
badsb(listerr, "INCOMPATIBLE FEATURE BITS IN SUPER BLOCK");
return 0;
}
if (memcmp(sblk.b_un.b_fs, asblk.b_un.b_fs, SBSIZE)) {
if (debug) {
u_int32_t *nlp, *olp, *endlp;
printf("superblock mismatches\n");
nlp = (u_int32_t *)asblk.b_un.b_fs;
olp = (u_int32_t *)sblk.b_un.b_fs;
endlp = olp + (SBSIZE / sizeof(*olp));
for ( ; olp < endlp; olp++, nlp++) {
if (*olp == *nlp)
continue;
printf("offset %ld, original %ld, "
"alternate %ld\n",
(long)(olp - (u_int32_t *)sblk.b_un.b_fs),
(long)fs2h32(*olp),
(long)fs2h32(*nlp));
}
}
badsb(listerr,
"VALUES IN SUPER BLOCK DISAGREE WITH "
"THOSE IN FIRST ALTERNATE");
return 0;
}
havesb = 1;
return 1;
}
int
ext2fs_bmaparray(struct vnode *vp,
#undef struct
daddr_t bn, daddr_t *bnp, struct indir *ap,
int *nump, int *runp)
{
struct inode *ip;
struct buf *bp, *cbp;
#define struct
// struct ufsmount *ump;
struct mount *mp;
#undef struct
struct indir a[NIADDR+1], *xap;
daddr_t daddr;
daddr_t metalbn;
int error, maxrun = 0, num;
ip = VTOI(vp);
mp = EXT2_SIMPLE_FILE_SYSTEM_PRIVATE_DATA_FROM_THIS(vp->Filesystem);
// mp = vp->v_mount; !!!! need to fix this badly!
// ump = ip->i_ump; NEED TO DO SOMETHING ABOUT ufsmount
#ifdef DIAGNOSTIC
if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL))
panic("ext2fs_bmaparray: invalid arguments");
#endif
if (runp) {
/*
* XXX
* If MAXBSIZE is the largest transfer the disks can handle,
* we probably want maxrun to be 1 block less so that we
* don't create a block larger than the device can handle.
*/
*runp = 0;
maxrun = MAXBSIZE / //mp->mnt_stat.f_iosize - 1; NEEDS FIX!!!
mp->fs->e2fs_bsize - 1;
}
if (bn >= 0 && bn < NDADDR) {
/* XXX ondisk32 */
*bnp = blkptrtodb(ump, fs2h32(ip->i_e2fs_blocks[bn]));
if (*bnp == 0)
*bnp = -1;
else if (runp)
/* XXX ondisk32 */
for (++bn; bn < NDADDR && *runp < maxrun &&
is_sequential(ump, (daddr_t)fs2h32(ip->i_e2fs_blocks[bn - 1]),
(daddr_t)fs2h32(ip->i_e2fs_blocks[bn]));
++bn, ++*runp);
return (0);
}
xap = ap == NULL ? a : ap;
if (!nump)
nump = #
if ((error = ufs_getlbns(vp, bn, xap, nump)) != 0)
return (error);
num = *nump;
/* Get disk address out of indirect block array */
/* XXX ondisk32 */
daddr = fs2h32(ip->i_e2fs_blocks[NDADDR + xap->in_off]);
#ifdef DIAGNOSTIC
if (num > NIADDR + 1 || num < 1) {
printf("ext2fs_bmaparray: num=%d\n", num);
panic("ext2fs_bmaparray: num");
}
#endif
for (bp = NULL, ++xap; --num; ++xap) {
/*
* Exit the loop if there is no disk address assigned yet and
* the indirect block isn't in the cache, or if we were
* looking for an indirect block and we've found it.
*/
metalbn = xap->in_lbn;
if (metalbn == bn)
break;
if (daddr == 0) {
mutex_enter(&bufcache_lock);
cbp = incore(vp, metalbn);
mutex_exit(&bufcache_lock);
if (cbp == NULL)
break;
}
/*
* If we get here, we've either got the block in the cache
* or we have a disk address for it, go fetch it.
*/
if (bp)
brelse(bp, 0);
xap->in_exists = 1;
//!!!!!!!!!!!!!!replaced 3rd param with 1 ftw
bp = getblk(vp, metalbn, 1, 0, 0);
if (bp == NULL) {
/*
* getblk() above returns NULL only iff we are
* pagedaemon. See the implementation of getblk
* for detail.
*/
return (ENOMEM);
}
if (bp->b_oflags & (BO_DONE | BO_DELWRI)) {
trace(TR_BREADHIT, pack(vp, size), metalbn);
}
#ifdef DIAGNOSTIC
else if (!daddr)
panic("ext2fs_bmaparry: indirect block not in cache");
#endif
else {
trace(TR_BREADMISS, pack(vp, size), metalbn);
bp->b_blkno = blkptrtodb(ump, daddr);
bp->b_flags |= B_READ;
VOP_STRATEGY(vp, bp);
// curlwp->l_ru.ru_inblock++; *//* XXX */
if ((error = biowait(bp)) != 0) {
brelse(bp, 0);
return (error);
}
}
/* XXX ondisk32 */
daddr = fs2h32(((int32_t *)bp->b_data)[xap->in_off]);
if (num == 1 && daddr && runp)
/* XXX ondisk32 */
for (bn = xap->in_off + 1;
bn < MNINDIR(ump) && *runp < maxrun &&
is_sequential(ump, ((int32_t *)bp->b_data)[bn - 1],
((int32_t *)bp->b_data)[bn]);
++bn, ++*runp);
}
if (bp)
brelse(bp, 0);
daddr = blkptrtodb(ump, daddr);
*bnp = daddr == 0 ? -1 : daddr;
return (0);
}
/*
* Balloc defines the structure of file system storage
* by allocating the physical blocks on a device given
* the inode and the logical block number in a file.
*/
int
ext2fs_balloc(struct inode *ip, daddr_t bn, int size,
kauth_cred_t cred, struct buf **bpp, int flags)
{
struct m_ext2fs *fs;
daddr_t nb;
struct buf *bp, *nbp;
struct vnode *vp = ITOV(ip);
struct indir indirs[EXT2FS_NIADDR + 2];
daddr_t newb, lbn, pref;
int32_t *bap; /* XXX ondisk32 */
int num, i, error;
u_int deallocated;
daddr_t *blkp, *allocblk, allociblk[EXT2FS_NIADDR + 1];
int32_t *allocib; /* XXX ondisk32 */
int unwindidx = -1;
UVMHIST_FUNC("ext2fs_balloc"); UVMHIST_CALLED(ubchist);
UVMHIST_LOG(ubchist, "bn 0x%x", bn,0,0,0);
if (bpp != NULL) {
*bpp = NULL;
}
if (bn < 0)
return (EFBIG);
fs = ip->i_e2fs;
lbn = bn;
/*
* The first EXT2FS_NDADDR blocks are direct blocks
*/
if (bn < EXT2FS_NDADDR) {
/* XXX ondisk32 */
nb = fs2h32(ip->i_e2fs_blocks[bn]);
if (nb != 0) {
/*
* the block is already allocated, just read it.
*/
if (bpp != NULL) {
error = bread(vp, bn, fs->e2fs_bsize, NOCRED,
B_MODIFY, &bp);
if (error) {
return (error);
}
*bpp = bp;
}
return (0);
}
/*
* allocate a new direct block.
*/
error = ext2fs_alloc(ip, bn,
ext2fs_blkpref(ip, bn, bn, &ip->i_e2fs_blocks[0]),
cred, &newb);
if (error)
return (error);
ip->i_e2fs_last_lblk = lbn;
ip->i_e2fs_last_blk = newb;
/* XXX ondisk32 */
ip->i_e2fs_blocks[bn] = h2fs32((int32_t)newb);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
if (bpp != NULL) {
bp = getblk(vp, bn, fs->e2fs_bsize, 0, 0);
bp->b_blkno = EXT2_FSBTODB(fs, newb);
if (flags & B_CLRBUF)
clrbuf(bp);
*bpp = bp;
}
return (0);
}
/*
* Determine the number of levels of indirection.
*/
pref = 0;
if ((error = ufs_getlbns(vp, bn, indirs, &num)) != 0)
return(error);
#ifdef DIAGNOSTIC
if (num < 1)
panic ("ext2fs_balloc: ufs_getlbns returned indirect block\n");
#endif
/*
* Fetch the first indirect block allocating if necessary.
*/
--num;
/* XXX ondisk32 */
nb = fs2h32(ip->i_e2fs_blocks[EXT2FS_NDADDR + indirs[0].in_off]);
allocib = NULL;
allocblk = allociblk;
if (nb == 0) {
pref = ext2fs_blkpref(ip, lbn, 0, (int32_t *)0);
error = ext2fs_alloc(ip, lbn, pref, cred, &newb);
if (error)
return (error);
nb = newb;
*allocblk++ = nb;
ip->i_e2fs_last_blk = newb;
bp = getblk(vp, indirs[1].in_lbn, fs->e2fs_bsize, 0, 0);
bp->b_blkno = EXT2_FSBTODB(fs, newb);
clrbuf(bp);
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if ((error = bwrite(bp)) != 0)
goto fail;
unwindidx = 0;
allocib = &ip->i_e2fs_blocks[EXT2FS_NDADDR + indirs[0].in_off];
/* XXX ondisk32 */
*allocib = h2fs32((int32_t)newb);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
}
/*
* Fetch through the indirect blocks, allocating as necessary.
*/
for (i = 1;;) {
error = bread(vp,
indirs[i].in_lbn, (int)fs->e2fs_bsize, NOCRED, 0, &bp);
if (error) {
goto fail;
}
bap = (int32_t *)bp->b_data; /* XXX ondisk32 */
nb = fs2h32(bap[indirs[i].in_off]);
if (i == num)
break;
i++;
if (nb != 0) {
brelse(bp, 0);
continue;
}
pref = ext2fs_blkpref(ip, lbn, 0, (int32_t *)0);
error = ext2fs_alloc(ip, lbn, pref, cred, &newb);
if (error) {
brelse(bp, 0);
goto fail;
}
nb = newb;
*allocblk++ = nb;
ip->i_e2fs_last_blk = newb;
nbp = getblk(vp, indirs[i].in_lbn, fs->e2fs_bsize, 0, 0);
nbp->b_blkno = EXT2_FSBTODB(fs, nb);
clrbuf(nbp);
/*
* Write synchronously so that indirect blocks
* never point at garbage.
*/
if ((error = bwrite(nbp)) != 0) {
brelse(bp, 0);
goto fail;
}
if (unwindidx < 0)
unwindidx = i - 1;
/* XXX ondisk32 */
bap[indirs[i - 1].in_off] = h2fs32((int32_t)nb);
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & B_SYNC) {
bwrite(bp);
} else {
bdwrite(bp);
}
}
/*
* Get the data block, allocating if necessary.
*/
if (nb == 0) {
pref = ext2fs_blkpref(ip, lbn, indirs[num].in_off, &bap[0]);
error = ext2fs_alloc(ip, lbn, pref, cred, &newb);
if (error) {
brelse(bp, 0);
goto fail;
}
nb = newb;
*allocblk++ = nb;
ip->i_e2fs_last_lblk = lbn;
ip->i_e2fs_last_blk = newb;
/* XXX ondisk32 */
bap[indirs[num].in_off] = h2fs32((int32_t)nb);
/*
* If required, write synchronously, otherwise use
* delayed write.
*/
if (flags & B_SYNC) {
bwrite(bp);
} else {
bdwrite(bp);
}
if (bpp != NULL) {
nbp = getblk(vp, lbn, fs->e2fs_bsize, 0, 0);
nbp->b_blkno = EXT2_FSBTODB(fs, nb);
if (flags & B_CLRBUF)
clrbuf(nbp);
*bpp = nbp;
}
return (0);
}
brelse(bp, 0);
if (bpp != NULL) {
if (flags & B_CLRBUF) {
error = bread(vp, lbn, (int)fs->e2fs_bsize, NOCRED,
B_MODIFY, &nbp);
if (error) {
goto fail;
}
} else {
nbp = getblk(vp, lbn, fs->e2fs_bsize, 0, 0);
nbp->b_blkno = EXT2_FSBTODB(fs, nb);
}
*bpp = nbp;
}
return (0);
fail:
/*
* If we have failed part way through block allocation, we
* have to deallocate any indirect blocks that we have allocated.
*/
for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
ext2fs_blkfree(ip, *blkp);
deallocated += fs->e2fs_bsize;
}
if (unwindidx >= 0) {
if (unwindidx == 0) {
*allocib = 0;
} else {
int r;
r = bread(vp, indirs[unwindidx].in_lbn,
(int)fs->e2fs_bsize, NOCRED, B_MODIFY, &bp);
if (r) {
panic("Could not unwind indirect block, error %d", r);
} else {
bap = (int32_t *)bp->b_data; /* XXX ondisk32 */
bap[indirs[unwindidx].in_off] = 0;
if (flags & B_SYNC)
bwrite(bp);
else
bdwrite(bp);
}
}
for (i = unwindidx + 1; i <= num; i++) {
bp = getblk(vp, indirs[i].in_lbn, (int)fs->e2fs_bsize,
0, 0);
brelse(bp, BC_INVAL);
}
}
if (deallocated) {
ext2fs_setnblock(ip, ext2fs_nblock(ip) - btodb(deallocated));
ip->i_e2fs_flags |= IN_CHANGE | IN_UPDATE;
}
return error;
}
