void jfs_set_inode_flags(struct inode *inode)
{
unsigned int flags = JFS_IP(inode)->mode2;
inode->i_flags &= ~(S_IMMUTABLE | S_APPEND |
S_NOATIME | S_DIRSYNC | S_SYNC);
if (flags & JFS_IMMUTABLE_FL)
inode->i_flags |= S_IMMUTABLE;
if (flags & JFS_APPEND_FL)
inode->i_flags |= S_APPEND;
if (flags & JFS_NOATIME_FL)
inode->i_flags |= S_NOATIME;
if (flags & JFS_DIRSYNC_FL)
inode->i_flags |= S_DIRSYNC;
if (flags & JFS_SYNC_FL)
inode->i_flags |= S_SYNC;
}
struct inode *jfs_iget(struct super_block *sb, unsigned long ino)
{
struct inode *inode;
int ret;
inode = iget_locked(sb, ino);
if (!inode)
return ERR_PTR(-ENOMEM);
if (!(inode->i_state & I_NEW))
return inode;
ret = diRead(inode);
if (ret < 0) {
iget_failed(inode);
return ERR_PTR(ret);
}
if (S_ISREG(inode->i_mode)) {
inode->i_op = &jfs_file_inode_operations;
inode->i_fop = &jfs_file_operations;
inode->i_mapping->a_ops = &jfs_aops;
} else if (S_ISDIR(inode->i_mode)) {
inode->i_op = &jfs_dir_inode_operations;
inode->i_fop = &jfs_dir_operations;
} else if (S_ISLNK(inode->i_mode)) {
if (inode->i_size >= IDATASIZE) {
inode->i_op = &page_symlink_inode_operations;
inode->i_mapping->a_ops = &jfs_aops;
} else {
inode->i_op = &jfs_fast_symlink_inode_operations;
/*
* The inline data should be null-terminated, but
* don't let on-disk corruption crash the kernel
*/
JFS_IP(inode)->i_inline[inode->i_size] = '\0';
}
} else {
inode->i_op = &jfs_file_inode_operations;
init_special_inode(inode, inode->i_mode, inode->i_rdev);
}
unlock_new_inode(inode);
return inode;
}
void jfs_clear_inode(struct inode *inode)
{
struct jfs_inode_info *ji = JFS_IP(inode);
if (is_bad_inode(inode))
/*
* We free the fs-dependent structure before making the
* inode bad
*/
return;
jfs_info("jfs_clear_inode called ip = 0x%p", inode);
if (ji->active_ag != -1) {
struct bmap *bmap = JFS_SBI(inode->i_sb)->bmap;
atomic_dec(&bmap->db_active[ji->active_ag]);
}
ASSERT(list_empty(&ji->anon_inode_list));
#ifdef CONFIG_JFS_POSIX_ACL
if (ji->i_acl != JFS_ACL_NOT_CACHED) {
posix_acl_release(ji->i_acl);
ji->i_acl = JFS_ACL_NOT_CACHED;
}
if (ji->i_default_acl != JFS_ACL_NOT_CACHED) {
posix_acl_release(ji->i_default_acl);
ji->i_default_acl = JFS_ACL_NOT_CACHED;
}
#endif
if (ji->atlhead) {
jfs_err("jfs_clear_inode: inode %p has anonymous tlocks",
inode);
jfs_err("i_state = 0x%lx, cflag = 0x%lx", inode->i_state,
ji->cflag);
}
free_jfs_inode(inode);
}
void jfs_delete_inode(struct inode *inode)
{
jfs_info("In jfs_delete_inode, inode = 0x%p", inode);
if (!is_bad_inode(inode) &&
(JFS_IP(inode)->fileset == FILESYSTEM_I)) {
truncate_inode_pages(&inode->i_data, 0);
if (test_cflag(COMMIT_Freewmap, inode))
jfs_free_zero_link(inode);
diFree(inode);
/*
* Free the inode from the quota allocation.
*/
DQUOT_INIT(inode);
DQUOT_FREE_INODE(inode);
DQUOT_DROP(inode);
}
clear_inode(inode);
}
/*
* jfs_permission()
*
* modified vfs_permission to check posix acl
*/
int jfs_permission(struct inode * inode, int mask, struct nameidata *nd)
{
umode_t mode = inode->i_mode;
struct jfs_inode_info *ji = JFS_IP(inode);
if (mask & MAY_WRITE) {
/*
* Nobody gets write access to a read-only fs.
*/
if (IS_RDONLY(inode) &&
(S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)))
return -EROFS;
/*
* Nobody gets write access to an immutable file.
*/
if (IS_IMMUTABLE(inode))
return -EACCES;
}
if (current->fsuid == inode->i_uid) {
mode >>= 6;
goto check_mode;
}
/*
* NAME: jfs_rename
*
* FUNCTION: rename a file or directory
*/
static int jfs_rename(struct inode *old_dir, struct dentry *old_dentry,
struct inode *new_dir, struct dentry *new_dentry)
{
struct btstack btstack;
ino_t ino;
struct component_name new_dname;
struct inode *new_ip;
struct component_name old_dname;
struct inode *old_ip;
int rc;
tid_t tid;
struct tlock *tlck;
struct dt_lock *dtlck;
struct lv *lv;
int ipcount;
struct inode *iplist[4];
struct tblock *tblk;
s64 new_size = 0;
int commit_flag;
jfs_info("jfs_rename: %s %s", old_dentry->d_name.name,
new_dentry->d_name.name);
dquot_initialize(old_dir);
dquot_initialize(new_dir);
old_ip = old_dentry->d_inode;
new_ip = new_dentry->d_inode;
if ((rc = get_UCSname(&old_dname, old_dentry)))
goto out1;
if ((rc = get_UCSname(&new_dname, new_dentry)))
goto out2;
/*
* Make sure source inode number is what we think it is
*/
rc = dtSearch(old_dir, &old_dname, &ino, &btstack, JFS_LOOKUP);
if (rc || (ino != old_ip->i_ino)) {
rc = -ENOENT;
goto out3;
}
/*
* Make sure dest inode number (if any) is what we think it is
*/
rc = dtSearch(new_dir, &new_dname, &ino, &btstack, JFS_LOOKUP);
if (!rc) {
if ((!new_ip) || (ino != new_ip->i_ino)) {
rc = -ESTALE;
goto out3;
}
} else if (rc != -ENOENT)
goto out3;
else if (new_ip) {
/* no entry exists, but one was expected */
rc = -ESTALE;
goto out3;
}
if (S_ISDIR(old_ip->i_mode)) {
if (new_ip) {
if (!dtEmpty(new_ip)) {
rc = -ENOTEMPTY;
goto out3;
}
} else if ((new_dir != old_dir) &&
(new_dir->i_nlink == JFS_LINK_MAX)) {
rc = -EMLINK;
goto out3;
}
} else if (new_ip) {
IWRITE_LOCK(new_ip, RDWRLOCK_NORMAL);
/* Init inode for quota operations. */
dquot_initialize(new_ip);
}
/*
* The real work starts here
*/
tid = txBegin(new_dir->i_sb, 0);
/*
* How do we know the locking is safe from deadlocks?
* The vfs does the hard part for us. Any time we are taking nested
* commit_mutexes, the vfs already has i_mutex held on the parent.
* Here, the vfs has already taken i_mutex on both old_dir and new_dir.
*/
mutex_lock_nested(&JFS_IP(new_dir)->commit_mutex, COMMIT_MUTEX_PARENT);
mutex_lock_nested(&JFS_IP(old_ip)->commit_mutex, COMMIT_MUTEX_CHILD);
if (old_dir != new_dir)
mutex_lock_nested(&JFS_IP(old_dir)->commit_mutex,
COMMIT_MUTEX_SECOND_PARENT);
if (new_ip) {
mutex_lock_nested(&JFS_IP(new_ip)->commit_mutex,
COMMIT_MUTEX_VICTIM);
/*
* Change existing directory entry to new inode number
*/
ino = new_ip->i_ino;
rc = dtModify(tid, new_dir, &new_dname, &ino,
old_ip->i_ino, JFS_RENAME);
if (rc)
goto out4;
drop_nlink(new_ip);
if (S_ISDIR(new_ip->i_mode)) {
drop_nlink(new_ip);
if (new_ip->i_nlink) {
mutex_unlock(&JFS_IP(new_ip)->commit_mutex);
if (old_dir != new_dir)
mutex_unlock(&JFS_IP(old_dir)->commit_mutex);
mutex_unlock(&JFS_IP(old_ip)->commit_mutex);
mutex_unlock(&JFS_IP(new_dir)->commit_mutex);
if (!S_ISDIR(old_ip->i_mode) && new_ip)
IWRITE_UNLOCK(new_ip);
jfs_error(new_ip->i_sb,
"jfs_rename: new_ip->i_nlink != 0");
return -EIO;
}
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_DELETE;
tblk->u.ip = new_ip;
} else if (new_ip->i_nlink == 0) {
assert(!test_cflag(COMMIT_Nolink, new_ip));
/* free block resources */
if ((new_size = commitZeroLink(tid, new_ip)) < 0) {
txAbort(tid, 1); /* Marks FS Dirty */
rc = new_size;
goto out4;
}
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_DELETE;
tblk->u.ip = new_ip;
} else {
new_ip->i_ctime = CURRENT_TIME;
mark_inode_dirty(new_ip);
}
} else {
/*
* Add new directory entry
*/
rc = dtSearch(new_dir, &new_dname, &ino, &btstack,
JFS_CREATE);
if (rc) {
jfs_err("jfs_rename didn't expect dtSearch to fail "
"w/rc = %d", rc);
goto out4;
}
ino = old_ip->i_ino;
rc = dtInsert(tid, new_dir, &new_dname, &ino, &btstack);
if (rc) {
if (rc == -EIO)
jfs_err("jfs_rename: dtInsert returned -EIO");
goto out4;
}
if (S_ISDIR(old_ip->i_mode))
inc_nlink(new_dir);
}
/*
* Remove old directory entry
*/
ino = old_ip->i_ino;
rc = dtDelete(tid, old_dir, &old_dname, &ino, JFS_REMOVE);
if (rc) {
jfs_err("jfs_rename did not expect dtDelete to return rc = %d",
rc);
txAbort(tid, 1); /* Marks Filesystem dirty */
goto out4;
}
if (S_ISDIR(old_ip->i_mode)) {
drop_nlink(old_dir);
if (old_dir != new_dir) {
/*
* Change inode number of parent for moved directory
*/
JFS_IP(old_ip)->i_dtroot.header.idotdot =
cpu_to_le32(new_dir->i_ino);
/* Linelock header of dtree */
tlck = txLock(tid, old_ip,
(struct metapage *) &JFS_IP(old_ip)->bxflag,
tlckDTREE | tlckBTROOT | tlckRELINK);
dtlck = (struct dt_lock *) & tlck->lock;
ASSERT(dtlck->index == 0);
lv = & dtlck->lv[0];
lv->offset = 0;
lv->length = 1;
dtlck->index++;
}
}
/*
* Update ctime on changed/moved inodes & mark dirty
*/
old_ip->i_ctime = CURRENT_TIME;
mark_inode_dirty(old_ip);
new_dir->i_ctime = new_dir->i_mtime = current_fs_time(new_dir->i_sb);
mark_inode_dirty(new_dir);
/* Build list of inodes modified by this transaction */
ipcount = 0;
iplist[ipcount++] = old_ip;
if (new_ip)
iplist[ipcount++] = new_ip;
iplist[ipcount++] = old_dir;
if (old_dir != new_dir) {
iplist[ipcount++] = new_dir;
old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME;
mark_inode_dirty(old_dir);
}
/*
* Incomplete truncate of file data can
* result in timing problems unless we synchronously commit the
* transaction.
*/
if (new_size)
commit_flag = COMMIT_SYNC;
else
commit_flag = 0;
rc = txCommit(tid, ipcount, iplist, commit_flag);
out4:
txEnd(tid);
if (new_ip)
mutex_unlock(&JFS_IP(new_ip)->commit_mutex);
if (old_dir != new_dir)
mutex_unlock(&JFS_IP(old_dir)->commit_mutex);
mutex_unlock(&JFS_IP(old_ip)->commit_mutex);
mutex_unlock(&JFS_IP(new_dir)->commit_mutex);
while (new_size && (rc == 0)) {
tid = txBegin(new_ip->i_sb, 0);
mutex_lock(&JFS_IP(new_ip)->commit_mutex);
new_size = xtTruncate_pmap(tid, new_ip, new_size);
if (new_size < 0) {
txAbort(tid, 1);
rc = new_size;
} else
rc = txCommit(tid, 1, &new_ip, COMMIT_SYNC);
txEnd(tid);
mutex_unlock(&JFS_IP(new_ip)->commit_mutex);
}
if (new_ip && (new_ip->i_nlink == 0))
set_cflag(COMMIT_Nolink, new_ip);
out3:
free_UCSname(&new_dname);
out2:
free_UCSname(&old_dname);
out1:
if (new_ip && !S_ISDIR(new_ip->i_mode))
IWRITE_UNLOCK(new_ip);
/*
* Truncating the directory index table is not guaranteed. It
* may need to be done iteratively
*/
if (test_cflag(COMMIT_Stale, old_dir)) {
if (old_dir->i_size > 1)
jfs_truncate_nolock(old_dir, 0);
clear_cflag(COMMIT_Stale, old_dir);
}
jfs_info("jfs_rename: returning %d", rc);
return rc;
}
/*
* NAME: jfs_mknod
*
* FUNCTION: Create a special file (device)
*/
static int jfs_mknod(struct inode *dir, struct dentry *dentry,
int mode, dev_t rdev)
{
struct jfs_inode_info *jfs_ip;
struct btstack btstack;
struct component_name dname;
ino_t ino;
struct inode *ip;
struct inode *iplist[2];
int rc;
tid_t tid;
struct tblock *tblk;
if (!new_valid_dev(rdev))
return -EINVAL;
jfs_info("jfs_mknod: %s", dentry->d_name.name);
dquot_initialize(dir);
if ((rc = get_UCSname(&dname, dentry)))
goto out;
ip = ialloc(dir, mode);
if (IS_ERR(ip)) {
rc = PTR_ERR(ip);
goto out1;
}
jfs_ip = JFS_IP(ip);
tid = txBegin(dir->i_sb, 0);
mutex_lock_nested(&JFS_IP(dir)->commit_mutex, COMMIT_MUTEX_PARENT);
mutex_lock_nested(&JFS_IP(ip)->commit_mutex, COMMIT_MUTEX_CHILD);
rc = jfs_init_acl(tid, ip, dir);
if (rc)
goto out3;
rc = jfs_init_security(tid, ip, dir, &dentry->d_name);
if (rc) {
txAbort(tid, 0);
goto out3;
}
if ((rc = dtSearch(dir, &dname, &ino, &btstack, JFS_CREATE))) {
txAbort(tid, 0);
goto out3;
}
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_CREATE;
tblk->ino = ip->i_ino;
tblk->u.ixpxd = JFS_IP(ip)->ixpxd;
ino = ip->i_ino;
if ((rc = dtInsert(tid, dir, &dname, &ino, &btstack))) {
txAbort(tid, 0);
goto out3;
}
ip->i_op = &jfs_file_inode_operations;
jfs_ip->dev = new_encode_dev(rdev);
init_special_inode(ip, ip->i_mode, rdev);
mark_inode_dirty(ip);
dir->i_ctime = dir->i_mtime = CURRENT_TIME;
mark_inode_dirty(dir);
iplist[0] = dir;
iplist[1] = ip;
rc = txCommit(tid, 2, iplist, 0);
out3:
txEnd(tid);
mutex_unlock(&JFS_IP(ip)->commit_mutex);
mutex_unlock(&JFS_IP(dir)->commit_mutex);
if (rc) {
free_ea_wmap(ip);
ip->i_nlink = 0;
unlock_new_inode(ip);
iput(ip);
} else {
d_instantiate(dentry, ip);
unlock_new_inode(ip);
}
out1:
free_UCSname(&dname);
out:
jfs_info("jfs_mknod: returning %d", rc);
return rc;
}
static void *jfs_follow_link(struct dentry *dentry, struct nameidata *nd)
{
char *s = JFS_IP(dentry->d_inode)->i_inline;
nd_set_link(nd, s);
return NULL;
}
/*
* NAME: jfs_link(vp, dvp, name, crp)
*
* FUNCTION: create a link to <vp> by the name = <name>
* in the parent directory <dvp>
*
* PARAMETER: vp - target object
* dvp - parent directory of new link
* name - name of new link to target object
* crp - credential
*
* RETURN: Errors from subroutines
*
* note:
* JFS does NOT support link() on directories (to prevent circular
* path in the directory hierarchy);
* EPERM: the target object is a directory, and either the caller
* does not have appropriate privileges or the implementation prohibits
* using link() on directories [XPG4.2].
*
* JFS does NOT support links between file systems:
* EXDEV: target object and new link are on different file systems and
* implementation does not support links between file systems [XPG4.2].
*/
int jfs_link(struct dentry *old_dentry,
struct inode *dir, struct dentry *dentry)
{
int rc;
tid_t tid;
struct inode *ip = old_dentry->d_inode;
ino_t ino;
struct component_name dname;
struct btstack btstack;
struct inode *iplist[2];
jFYI(1,
("jfs_link: %s %s\n", old_dentry->d_name.name,
dentry->d_name.name));
/* JFS does NOT support link() on directories */
if (S_ISDIR(ip->i_mode))
return -EPERM;
tid = txBegin(ip->i_sb, 0);
down(&JFS_IP(dir)->commit_sem);
down(&JFS_IP(ip)->commit_sem);
if (ip->i_nlink == JFS_LINK_MAX) {
rc = EMLINK;
goto out;
}
/*
* scan parent directory for entry/freespace
*/
if ((rc = get_UCSname(&dname, dentry, JFS_SBI(ip->i_sb)->nls_tab)))
goto out;
if ((rc = dtSearch(dir, &dname, &ino, &btstack, JFS_CREATE)))
goto out;
/*
* create entry for new link in parent directory
*/
ino = ip->i_ino;
if ((rc = dtInsert(tid, dir, &dname, &ino, &btstack)))
goto out;
/* update object inode */
ip->i_nlink++; /* for new link */
ip->i_ctime = CURRENT_TIME;
mark_inode_dirty(dir);
atomic_inc(&ip->i_count);
d_instantiate(dentry, ip);
iplist[0] = ip;
iplist[1] = dir;
rc = txCommit(tid, 2, &iplist[0], 0);
out:
txEnd(tid);
up(&JFS_IP(dir)->commit_sem);
up(&JFS_IP(ip)->commit_sem);
jFYI(1, ("jfs_link: rc:%d\n", rc));
return -rc;
}
/*
* NAME: jfs_link(vp, dvp, name, crp)
*
* FUNCTION: create a link to <vp> by the name = <name>
* in the parent directory <dvp>
*
* PARAMETER: vp - target object
* dvp - parent directory of new link
* name - name of new link to target object
* crp - credential
*
* RETURN: Errors from subroutines
*
* note:
* JFS does NOT support link() on directories (to prevent circular
* path in the directory hierarchy);
* EPERM: the target object is a directory, and either the caller
* does not have appropriate privileges or the implementation prohibits
* using link() on directories [XPG4.2].
*
* JFS does NOT support links between file systems:
* EXDEV: target object and new link are on different file systems and
* implementation does not support links between file systems [XPG4.2].
*/
static int jfs_link(struct dentry *old_dentry,
struct inode *dir, struct dentry *dentry)
{
int rc;
tid_t tid;
struct inode *ip = old_dentry->d_inode;
ino_t ino;
struct component_name dname;
struct btstack btstack;
struct inode *iplist[2];
jfs_info("jfs_link: %s %s", old_dentry->d_name.name,
dentry->d_name.name);
if (ip->i_nlink == JFS_LINK_MAX)
return -EMLINK;
dquot_initialize(dir);
tid = txBegin(ip->i_sb, 0);
mutex_lock_nested(&JFS_IP(dir)->commit_mutex, COMMIT_MUTEX_PARENT);
mutex_lock_nested(&JFS_IP(ip)->commit_mutex, COMMIT_MUTEX_CHILD);
/*
* scan parent directory for entry/freespace
*/
if ((rc = get_UCSname(&dname, dentry)))
goto out;
if ((rc = dtSearch(dir, &dname, &ino, &btstack, JFS_CREATE)))
goto free_dname;
/*
* create entry for new link in parent directory
*/
ino = ip->i_ino;
if ((rc = dtInsert(tid, dir, &dname, &ino, &btstack)))
goto free_dname;
/* update object inode */
inc_nlink(ip); /* for new link */
ip->i_ctime = CURRENT_TIME;
dir->i_ctime = dir->i_mtime = CURRENT_TIME;
mark_inode_dirty(dir);
ihold(ip);
iplist[0] = ip;
iplist[1] = dir;
rc = txCommit(tid, 2, &iplist[0], 0);
if (rc) {
ip->i_nlink--; /* never instantiated */
iput(ip);
} else
d_instantiate(dentry, ip);
free_dname:
free_UCSname(&dname);
out:
txEnd(tid);
mutex_unlock(&JFS_IP(ip)->commit_mutex);
mutex_unlock(&JFS_IP(dir)->commit_mutex);
jfs_info("jfs_link: rc:%d", rc);
return rc;
}
long jfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
{
struct inode *inode = filp->f_dentry->d_inode;
struct jfs_inode_info *jfs_inode = JFS_IP(inode);
unsigned int flags;
switch (cmd) {
case JFS_IOC_GETFLAGS:
jfs_get_inode_flags(jfs_inode);
flags = jfs_inode->mode2 & JFS_FL_USER_VISIBLE;
flags = jfs_map_ext2(flags, 0);
return put_user(flags, (int __user *) arg);
case JFS_IOC_SETFLAGS: {
unsigned int oldflags;
int err;
err = mnt_want_write_file(filp);
if (err)
return err;
if (!inode_owner_or_capable(inode)) {
err = -EACCES;
goto setflags_out;
}
if (get_user(flags, (int __user *) arg)) {
err = -EFAULT;
goto setflags_out;
}
flags = jfs_map_ext2(flags, 1);
if (!S_ISDIR(inode->i_mode))
flags &= ~JFS_DIRSYNC_FL;
if (IS_NOQUOTA(inode)) {
err = -EPERM;
goto setflags_out;
}
mutex_lock(&inode->i_mutex);
jfs_get_inode_flags(jfs_inode);
oldflags = jfs_inode->mode2;
if ((oldflags & JFS_IMMUTABLE_FL) ||
((flags ^ oldflags) &
(JFS_APPEND_FL | JFS_IMMUTABLE_FL))) {
if (!capable(CAP_LINUX_IMMUTABLE)) {
mutex_unlock(&inode->i_mutex);
err = -EPERM;
goto setflags_out;
}
}
flags = flags & JFS_FL_USER_MODIFIABLE;
flags |= oldflags & ~JFS_FL_USER_MODIFIABLE;
jfs_inode->mode2 = flags;
jfs_set_inode_flags(inode);
mutex_unlock(&inode->i_mutex);
inode->i_ctime = CURRENT_TIME_SEC;
mark_inode_dirty(inode);
setflags_out:
mnt_drop_write_file(filp);
return err;
}
default:
return -ENOTTY;
}
}
/*
* NAME: ialloc()
*
* FUNCTION: Allocate a new inode
*
*/
struct inode *ialloc(struct inode *parent, umode_t mode)
{
struct super_block *sb = parent->i_sb;
struct inode *inode;
struct jfs_inode_info *jfs_inode;
int rc;
inode = new_inode(sb);
if (!inode) {
jfs_warn("ialloc: new_inode returned NULL!");
rc = -ENOMEM;
goto fail;
}
jfs_inode = JFS_IP(inode);
rc = diAlloc(parent, S_ISDIR(mode), inode);
if (rc) {
jfs_warn("ialloc: diAlloc returned %d!", rc);
if (rc == -EIO)
make_bad_inode(inode);
goto fail_put;
}
if (insert_inode_locked(inode) < 0) {
rc = -EINVAL;
goto fail_unlock;
}
inode_init_owner(inode, parent, mode);
/*
* New inodes need to save sane values on disk when
* uid & gid mount options are used
*/
jfs_inode->saved_uid = inode->i_uid;
jfs_inode->saved_gid = inode->i_gid;
/*
* Allocate inode to quota.
*/
dquot_initialize(inode);
rc = dquot_alloc_inode(inode);
if (rc)
goto fail_drop;
/* inherit flags from parent */
jfs_inode->mode2 = JFS_IP(parent)->mode2 & JFS_FL_INHERIT;
if (S_ISDIR(mode)) {
jfs_inode->mode2 |= IDIRECTORY;
jfs_inode->mode2 &= ~JFS_DIRSYNC_FL;
}
else {
jfs_inode->mode2 |= INLINEEA | ISPARSE;
if (S_ISLNK(mode))
jfs_inode->mode2 &= ~(JFS_IMMUTABLE_FL|JFS_APPEND_FL);
}
jfs_inode->mode2 |= inode->i_mode;
inode->i_blocks = 0;
inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
jfs_inode->otime = inode->i_ctime.tv_sec;
inode->i_generation = JFS_SBI(sb)->gengen++;
jfs_inode->cflag = 0;
/* Zero remaining fields */
memset(&jfs_inode->acl, 0, sizeof(dxd_t));
memset(&jfs_inode->ea, 0, sizeof(dxd_t));
jfs_inode->next_index = 0;
jfs_inode->acltype = 0;
jfs_inode->btorder = 0;
jfs_inode->btindex = 0;
jfs_inode->bxflag = 0;
jfs_inode->blid = 0;
jfs_inode->atlhead = 0;
jfs_inode->atltail = 0;
jfs_inode->xtlid = 0;
jfs_set_inode_flags(inode);
jfs_info("ialloc returns inode = 0x%p\n", inode);
return inode;
fail_drop:
dquot_drop(inode);
inode->i_flags |= S_NOQUOTA;
fail_unlock:
inode->i_nlink = 0;
unlock_new_inode(inode);
fail_put:
iput(inode);
fail:
return ERR_PTR(rc);
}
/*
* NAME: jfs_free_zero_link()
*
* FUNCTION: for non-directory, called by iClose(),
* free resources of a file from cache and WORKING map
* for a file previously committed with zero link count
* while associated with a pager object,
*
* PARAMETER: ip - pointer to inode of file.
*/
void jfs_free_zero_link(struct inode *ip)
{
int type;
jfs_info("jfs_free_zero_link: ip = 0x%p", ip);
/* return if not reg or symbolic link or if size is
* already ok.
*/
type = ip->i_mode & S_IFMT;
switch (type) {
case S_IFREG:
break;
case S_IFLNK:
/* if its contained in inode nothing to do */
if (ip->i_size < IDATASIZE)
return;
break;
default:
return;
}
/*
* free EA
*/
if (JFS_IP(ip)->ea.flag & DXD_EXTENT) {
s64 xaddr = addressDXD(&JFS_IP(ip)->ea);
int xlen = lengthDXD(&JFS_IP(ip)->ea);
struct maplock maplock; /* maplock for COMMIT_WMAP */
struct pxd_lock *pxdlock; /* maplock for COMMIT_WMAP */
/* free EA pages from cache */
invalidate_dxd_metapages(ip, JFS_IP(ip)->ea);
/* free EA extent from working block map */
maplock.index = 1;
pxdlock = (struct pxd_lock *) & maplock;
pxdlock->flag = mlckFREEPXD;
PXDaddress(&pxdlock->pxd, xaddr);
PXDlength(&pxdlock->pxd, xlen);
txFreeMap(ip, pxdlock, NULL, COMMIT_WMAP);
}
/*
* free ACL
*/
if (JFS_IP(ip)->acl.flag & DXD_EXTENT) {
s64 xaddr = addressDXD(&JFS_IP(ip)->acl);
int xlen = lengthDXD(&JFS_IP(ip)->acl);
struct maplock maplock; /* maplock for COMMIT_WMAP */
struct pxd_lock *pxdlock; /* maplock for COMMIT_WMAP */
invalidate_dxd_metapages(ip, JFS_IP(ip)->acl);
/* free ACL extent from working block map */
maplock.index = 1;
pxdlock = (struct pxd_lock *) & maplock;
pxdlock->flag = mlckFREEPXD;
PXDaddress(&pxdlock->pxd, xaddr);
PXDlength(&pxdlock->pxd, xlen);
txFreeMap(ip, pxdlock, NULL, COMMIT_WMAP);
}
/*
* free xtree/data (truncate to zero length):
* free xtree/data pages from cache, and
* free xtree/data blocks from working block map;
*/
if (ip->i_size)
xtTruncate(0, ip, 0, COMMIT_WMAP);
}
/*
* NAME: ialloc()
*
* FUNCTION: Allocate a new inode
*
*/
struct inode *ialloc(struct inode *parent, umode_t mode)
{
struct super_block *sb = parent->i_sb;
struct inode *inode;
struct jfs_inode_info *jfs_inode;
int rc;
inode = new_inode(sb);
if (!inode) {
jERROR(1, ("ialloc: new_inode returned NULL!\n"));
return inode;
}
rc = alloc_jfs_inode(inode);
if (rc) {
make_bad_inode(inode);
iput(inode);
return NULL;
}
jfs_inode = JFS_IP(inode);
rc = diAlloc(parent, S_ISDIR(mode), inode);
if (rc) {
jERROR(1, ("ialloc: diAlloc returned %d!\n", rc));
free_jfs_inode(inode);
make_bad_inode(inode);
iput(inode);
return NULL;
}
inode->i_uid = current->fsuid;
if (parent->i_mode & S_ISGID) {
inode->i_gid = parent->i_gid;
if (S_ISDIR(mode))
mode |= S_ISGID;
} else
inode->i_gid = current->fsgid;
inode->i_mode = mode;
if (S_ISDIR(mode))
jfs_inode->mode2 = IDIRECTORY | mode;
else
jfs_inode->mode2 = INLINEEA | ISPARSE | mode;
inode->i_blksize = sb->s_blocksize;
inode->i_blocks = 0;
inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
jfs_inode->otime = inode->i_ctime;
inode->i_generation = JFS_SBI(sb)->gengen++;
jfs_inode->cflag = 0;
set_cflag(COMMIT_New, inode);
/* Zero remaining fields */
memset(&jfs_inode->acl, 0, sizeof(dxd_t));
memset(&jfs_inode->ea, 0, sizeof(dxd_t));
jfs_inode->next_index = 0;
jfs_inode->acltype = 0;
jfs_inode->btorder = 0;
jfs_inode->btindex = 0;
jfs_inode->bxflag = 0;
jfs_inode->blid = 0;
jfs_inode->atlhead = 0;
jfs_inode->atltail = 0;
jfs_inode->xtlid = 0;
jFYI(1, ("ialloc returns inode = 0x%p\n", inode));
return inode;
}
/*
* NAME: jfs_rmdir(dip, dentry)
*
* FUNCTION: remove a link to child directory
*
* PARAMETER: dip - parent inode
* dentry - child directory dentry
*
* RETURN: EINVAL - if name is . or ..
* EINVAL - if . or .. exist but are invalid.
* errors from subroutines
*
* note:
* if other threads have the directory open when the last link
* is removed, the "." and ".." entries, if present, are removed before
* rmdir() returns and no new entries may be created in the directory,
* but the directory is not removed until the last reference to
* the directory is released (cf.unlink() of regular file).
*/
int jfs_rmdir(struct inode *dip, struct dentry *dentry)
{
int rc;
tid_t tid; /* transaction id */
struct inode *ip = dentry->d_inode;
ino_t ino;
struct component_name dname;
struct inode *iplist[2];
struct tblock *tblk;
jFYI(1, ("jfs_rmdir: dip:0x%p name:%s\n", dip, dentry->d_name.name));
/* directory must be empty to be removed */
if (!dtEmpty(ip)) {
rc = ENOTEMPTY;
goto out;
}
if ((rc = get_UCSname(&dname, dentry, JFS_SBI(dip->i_sb)->nls_tab))) {
goto out;
}
tid = txBegin(dip->i_sb, 0);
down(&JFS_IP(dip)->commit_sem);
down(&JFS_IP(ip)->commit_sem);
iplist[0] = dip;
iplist[1] = ip;
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_DELETE;
tblk->ip = ip;
/*
* delete the entry of target directory from parent directory
*/
ino = ip->i_ino;
if ((rc = dtDelete(tid, dip, &dname, &ino, JFS_REMOVE))) {
jERROR(1, ("jfs_rmdir: dtDelete returned %d\n", rc));
if (rc == EIO)
txAbort(tid, 1);
txEnd(tid);
up(&JFS_IP(dip)->commit_sem);
up(&JFS_IP(ip)->commit_sem);
goto out2;
}
/* update parent directory's link count corresponding
* to ".." entry of the target directory deleted
*/
dip->i_nlink--;
dip->i_ctime = dip->i_mtime = CURRENT_TIME;
mark_inode_dirty(dip);
/*
* OS/2 could have created EA and/or ACL
*/
/* free EA from both persistent and working map */
if (JFS_IP(ip)->ea.flag & DXD_EXTENT) {
/* free EA pages */
txEA(tid, ip, &JFS_IP(ip)->ea, NULL);
}
JFS_IP(ip)->ea.flag = 0;
/* free ACL from both persistent and working map */
if (JFS_IP(ip)->acl.flag & DXD_EXTENT) {
/* free ACL pages */
txEA(tid, ip, &JFS_IP(ip)->acl, NULL);
}
JFS_IP(ip)->acl.flag = 0;
/* mark the target directory as deleted */
ip->i_nlink = 0;
mark_inode_dirty(ip);
rc = txCommit(tid, 2, &iplist[0], 0);
txEnd(tid);
up(&JFS_IP(dip)->commit_sem);
up(&JFS_IP(ip)->commit_sem);
/*
* Truncating the directory index table is not guaranteed. It
* may need to be done iteratively
*/
if (test_cflag(COMMIT_Stale, dip)) {
if (dip->i_size > 1)
jfs_truncate_nolock(dip, 0);
clear_cflag(COMMIT_Stale, dip);
}
out2:
free_UCSname(&dname);
out:
jFYI(1, ("jfs_rmdir: rc:%d\n", rc));
return -rc;
}
/*
* NAME: jfs_mkdir(dip, dentry, mode)
*
* FUNCTION: create a child directory in the parent directory <dip>
* with name = <from dentry> and mode = <mode>
*
* PARAMETER: dip - parent directory vnode
* dentry - dentry of child directory
* mode - create mode (rwxrwxrwx).
*
* RETURN: Errors from subroutines
*
* note:
* EACCESS: user needs search+write permission on the parent directory
*/
static int jfs_mkdir(struct inode *dip, struct dentry *dentry, int mode)
{
int rc = 0;
tid_t tid; /* transaction id */
struct inode *ip = NULL; /* child directory inode */
ino_t ino;
struct component_name dname; /* child directory name */
struct btstack btstack;
struct inode *iplist[2];
struct tblock *tblk;
jfs_info("jfs_mkdir: dip:0x%p name:%s", dip, dentry->d_name.name);
/* link count overflow on parent directory ? */
if (dip->i_nlink == JFS_LINK_MAX) {
rc = -EMLINK;
goto out1;
}
/*
* search parent directory for entry/freespace
* (dtSearch() returns parent directory page pinned)
*/
if ((rc = get_UCSname(&dname, dentry)))
goto out1;
/*
* Either iAlloc() or txBegin() may block. Deadlock can occur if we
* block there while holding dtree page, so we allocate the inode &
* begin the transaction before we search the directory.
*/
ip = ialloc(dip, S_IFDIR | mode);
if (ip == NULL) {
rc = -ENOSPC;
goto out2;
}
tid = txBegin(dip->i_sb, 0);
mutex_lock(&JFS_IP(dip)->commit_mutex);
mutex_lock(&JFS_IP(ip)->commit_mutex);
rc = jfs_init_acl(tid, ip, dip);
if (rc)
goto out3;
rc = jfs_init_security(tid, ip, dip);
if (rc) {
txAbort(tid, 0);
goto out3;
}
if ((rc = dtSearch(dip, &dname, &ino, &btstack, JFS_CREATE))) {
jfs_err("jfs_mkdir: dtSearch returned %d", rc);
txAbort(tid, 0);
goto out3;
}
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_CREATE;
tblk->ino = ip->i_ino;
tblk->u.ixpxd = JFS_IP(ip)->ixpxd;
iplist[0] = dip;
iplist[1] = ip;
/*
* initialize the child directory in-line in inode
*/
dtInitRoot(tid, ip, dip->i_ino);
/*
* create entry in parent directory for child directory
* (dtInsert() releases parent directory page)
*/
ino = ip->i_ino;
if ((rc = dtInsert(tid, dip, &dname, &ino, &btstack))) {
if (rc == -EIO) {
jfs_err("jfs_mkdir: dtInsert returned -EIO");
txAbort(tid, 1); /* Marks Filesystem dirty */
} else
txAbort(tid, 0); /* Filesystem full */
goto out3;
}
ip->i_nlink = 2; /* for '.' */
ip->i_op = &jfs_dir_inode_operations;
ip->i_fop = &jfs_dir_operations;
insert_inode_hash(ip);
mark_inode_dirty(ip);
/* update parent directory inode */
dip->i_nlink++; /* for '..' from child directory */
dip->i_ctime = dip->i_mtime = CURRENT_TIME;
mark_inode_dirty(dip);
rc = txCommit(tid, 2, &iplist[0], 0);
out3:
txEnd(tid);
mutex_unlock(&JFS_IP(ip)->commit_mutex);
mutex_unlock(&JFS_IP(dip)->commit_mutex);
if (rc) {
free_ea_wmap(ip);
ip->i_nlink = 0;
iput(ip);
} else
d_instantiate(dentry, ip);
out2:
free_UCSname(&dname);
out1:
jfs_info("jfs_mkdir: rc:%d", rc);
return rc;
}
/*
* NAME: jfs_unlink(dip, dentry)
*
* FUNCTION: remove a link to object <vp> named by <name>
* from parent directory <dvp>
*
* PARAMETER: dip - inode of parent directory
* dentry - dentry of object to be removed
*
* RETURN: errors from subroutines
*
* note:
* temporary file: if one or more processes have the file open
* when the last link is removed, the link will be removed before
* unlink() returns, but the removal of the file contents will be
* postponed until all references to the files are closed.
*
* JFS does NOT support unlink() on directories.
*
*/
int jfs_unlink(struct inode *dip, struct dentry *dentry)
{
int rc;
tid_t tid; /* transaction id */
struct inode *ip = dentry->d_inode;
ino_t ino;
struct component_name dname; /* object name */
struct inode *iplist[2];
struct tblock *tblk;
s64 new_size = 0;
int commit_flag;
jFYI(1, ("jfs_unlink: dip:0x%p name:%s\n", dip, dentry->d_name.name));
if ((rc = get_UCSname(&dname, dentry, JFS_SBI(dip->i_sb)->nls_tab)))
goto out;
IWRITE_LOCK(ip);
tid = txBegin(dip->i_sb, 0);
down(&JFS_IP(dip)->commit_sem);
down(&JFS_IP(ip)->commit_sem);
iplist[0] = dip;
iplist[1] = ip;
/*
* delete the entry of target file from parent directory
*/
ino = ip->i_ino;
if ((rc = dtDelete(tid, dip, &dname, &ino, JFS_REMOVE))) {
jERROR(1, ("jfs_unlink: dtDelete returned %d\n", rc));
if (rc == EIO)
txAbort(tid, 1); /* Marks FS Dirty */
txEnd(tid);
up(&JFS_IP(dip)->commit_sem);
up(&JFS_IP(ip)->commit_sem);
IWRITE_UNLOCK(ip);
goto out1;
}
ASSERT(ip->i_nlink);
ip->i_ctime = dip->i_ctime = dip->i_mtime = CURRENT_TIME;
mark_inode_dirty(dip);
/* update target's inode */
ip->i_nlink--;
mark_inode_dirty(ip);
/*
* commit zero link count object
*/
if (ip->i_nlink == 0) {
assert(!test_cflag(COMMIT_Nolink, ip));
/* free block resources */
if ((new_size = commitZeroLink(tid, ip)) < 0) {
txAbort(tid, 1); /* Marks FS Dirty */
txEnd(tid);
up(&JFS_IP(dip)->commit_sem);
up(&JFS_IP(ip)->commit_sem);
IWRITE_UNLOCK(ip);
rc = -new_size; /* We return -rc */
goto out1;
}
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_DELETE;
tblk->ip = ip;
}
/*
* Incomplete truncate of file data can
* result in timing problems unless we synchronously commit the
* transaction.
*/
if (new_size)
commit_flag = COMMIT_SYNC;
else
commit_flag = 0;
/*
* If xtTruncate was incomplete, commit synchronously to avoid
* timing complications
*/
rc = txCommit(tid, 2, &iplist[0], commit_flag);
txEnd(tid);
up(&JFS_IP(dip)->commit_sem);
up(&JFS_IP(ip)->commit_sem);
while (new_size && (rc == 0)) {
tid = txBegin(dip->i_sb, 0);
down(&JFS_IP(ip)->commit_sem);
new_size = xtTruncate_pmap(tid, ip, new_size);
if (new_size < 0) {
txAbort(tid, 1); /* Marks FS Dirty */
rc = -new_size; /* We return -rc */
} else
rc = txCommit(tid, 2, &iplist[0], COMMIT_SYNC);
txEnd(tid);
up(&JFS_IP(ip)->commit_sem);
}
if (ip->i_nlink == 0)
set_cflag(COMMIT_Nolink, ip);
if (!test_cflag(COMMIT_Holdlock, ip))
IWRITE_UNLOCK(ip);
/*
* Truncating the directory index table is not guaranteed. It
* may need to be done iteratively
*/
if (test_cflag(COMMIT_Stale, dip)) {
if (dip->i_size > 1)
jfs_truncate_nolock(dip, 0);
clear_cflag(COMMIT_Stale, dip);
}
out1:
free_UCSname(&dname);
out:
jFYI(1, ("jfs_unlink: rc:%d\n", -rc));
return -rc;
}
int jfs_symlink(struct inode *dip, struct dentry *dentry, const char *name)
{
int rc;
tid_t tid;
ino_t ino = 0;
struct component_name dname;
int ssize; /* source pathname size */
struct btstack btstack;
struct inode *ip = dentry->d_inode;
unchar *i_fastsymlink;
s64 xlen = 0;
int bmask = 0, xsize;
s64 xaddr;
struct metapage *mp;
struct super_block *sb;
struct tblock *tblk;
struct inode *iplist[2];
jFYI(1, ("jfs_symlink: dip:0x%p name:%s\n", dip, name));
ssize = strlen(name) + 1;
/*
* search parent directory for entry/freespace
* (dtSearch() returns parent directory page pinned)
*/
if ((rc = get_UCSname(&dname, dentry, JFS_SBI(dip->i_sb)->nls_tab)))
goto out1;
/*
* allocate on-disk/in-memory inode for symbolic link:
* (iAlloc() returns new, locked inode)
*/
ip = ialloc(dip, S_IFLNK | 0777);
if (ip == NULL) {
rc = ENOSPC;
goto out2;
}
tid = txBegin(dip->i_sb, 0);
down(&JFS_IP(dip)->commit_sem);
down(&JFS_IP(ip)->commit_sem);
if ((rc = dtSearch(dip, &dname, &ino, &btstack, JFS_CREATE)))
goto out3;
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_CREATE;
tblk->ip = ip;
/*
* create entry for symbolic link in parent directory
*/
ino = ip->i_ino;
if ((rc = dtInsert(tid, dip, &dname, &ino, &btstack))) {
jERROR(1, ("jfs_symlink: dtInsert returned %d\n", rc));
/* discard ne inode */
goto out3;
}
/* fix symlink access permission
* (dir_create() ANDs in the u.u_cmask,
* but symlinks really need to be 777 access)
*/
ip->i_mode |= 0777;
/*
* write symbolic link target path name
*/
xtInitRoot(tid, ip);
/*
* write source path name inline in on-disk inode (fast symbolic link)
*/
if (ssize <= IDATASIZE) {
ip->i_op = &jfs_symlink_inode_operations;
i_fastsymlink = JFS_IP(ip)->i_inline;
memcpy(i_fastsymlink, name, ssize);
ip->i_size = ssize - 1;
/*
* if symlink is > 128 bytes, we don't have the space to
* store inline extended attributes
*/
if (ssize > sizeof (JFS_IP(ip)->i_inline))
JFS_IP(ip)->mode2 &= ~INLINEEA;
jFYI(1,
("jfs_symlink: fast symlink added ssize:%d name:%s \n",
ssize, name));
}
/*
* write source path name in a single extent
*/
else {
jFYI(1, ("jfs_symlink: allocate extent ip:0x%p\n", ip));
ip->i_op = &page_symlink_inode_operations;
ip->i_mapping->a_ops = &jfs_aops;
/*
* even though the data of symlink object (source
* path name) is treated as non-journaled user data,
* it is read/written thru buffer cache for performance.
*/
sb = ip->i_sb;
bmask = JFS_SBI(sb)->bsize - 1;
xsize = (ssize + bmask) & ~bmask;
xaddr = 0;
xlen = xsize >> JFS_SBI(sb)->l2bsize;
if ((rc = xtInsert(tid, ip, 0, 0, xlen, &xaddr, 0)) == 0) {
ip->i_size = ssize - 1;
while (ssize) {
int copy_size = min(ssize, PSIZE);
mp = get_metapage(ip, xaddr, PSIZE, 1);
if (mp == NULL) {
dtDelete(tid, dip, &dname, &ino,
JFS_REMOVE);
rc = EIO;
goto out3;
}
memcpy(mp->data, name, copy_size);
flush_metapage(mp);
#if 0
mark_buffer_uptodate(bp, 1);
mark_buffer_dirty(bp, 1);
if (IS_SYNC(dip)) {
ll_rw_block(WRITE, 1, &bp);
wait_on_buffer(bp);
}
brelse(bp);
#endif /* 0 */
ssize -= copy_size;
xaddr += JFS_SBI(sb)->nbperpage;
}
ip->i_blocks = LBLK2PBLK(sb, xlen);
} else {
dtDelete(tid, dip, &dname, &ino, JFS_REMOVE);
rc = ENOSPC;
goto out3;
}
}
insert_inode_hash(ip);
mark_inode_dirty(ip);
d_instantiate(dentry, ip);
/*
* commit update of parent directory and link object
*
* if extent allocation failed (ENOSPC),
* the parent inode is committed regardless to avoid
* backing out parent directory update (by dtInsert())
* and subsequent dtDelete() which is harmless wrt
* integrity concern.
* the symlink inode will be freed by iput() at exit
* as it has a zero link count (by dtDelete()) and
* no permanant resources.
*/
iplist[0] = dip;
if (rc == 0) {
iplist[1] = ip;
rc = txCommit(tid, 2, &iplist[0], 0);
} else
rc = txCommit(tid, 1, &iplist[0], 0);
out3:
txEnd(tid);
up(&JFS_IP(dip)->commit_sem);
up(&JFS_IP(ip)->commit_sem);
if (rc) {
ip->i_nlink = 0;
iput(ip);
}
out2:
free_UCSname(&dname);
out1:
jFYI(1, ("jfs_symlink: rc:%d\n", -rc));
return -rc;
}
/*
* NAME: jfs_rmdir(dip, dentry)
*
* FUNCTION: remove a link to child directory
*
* PARAMETER: dip - parent inode
* dentry - child directory dentry
*
* RETURN: -EINVAL - if name is . or ..
* -EINVAL - if . or .. exist but are invalid.
* errors from subroutines
*
* note:
* if other threads have the directory open when the last link
* is removed, the "." and ".." entries, if present, are removed before
* rmdir() returns and no new entries may be created in the directory,
* but the directory is not removed until the last reference to
* the directory is released (cf.unlink() of regular file).
*/
static int jfs_rmdir(struct inode *dip, struct dentry *dentry)
{
int rc;
tid_t tid; /* transaction id */
struct inode *ip = dentry->d_inode;
ino_t ino;
struct component_name dname;
struct inode *iplist[2];
struct tblock *tblk;
jfs_info("jfs_rmdir: dip:0x%p name:%s", dip, dentry->d_name.name);
/* Init inode for quota operations. */
dquot_initialize(dip);
dquot_initialize(ip);
/* directory must be empty to be removed */
if (!dtEmpty(ip)) {
rc = -ENOTEMPTY;
goto out;
}
if ((rc = get_UCSname(&dname, dentry))) {
goto out;
}
tid = txBegin(dip->i_sb, 0);
mutex_lock_nested(&JFS_IP(dip)->commit_mutex, COMMIT_MUTEX_PARENT);
mutex_lock_nested(&JFS_IP(ip)->commit_mutex, COMMIT_MUTEX_CHILD);
iplist[0] = dip;
iplist[1] = ip;
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_DELETE;
tblk->u.ip = ip;
/*
* delete the entry of target directory from parent directory
*/
ino = ip->i_ino;
if ((rc = dtDelete(tid, dip, &dname, &ino, JFS_REMOVE))) {
jfs_err("jfs_rmdir: dtDelete returned %d", rc);
if (rc == -EIO)
txAbort(tid, 1);
txEnd(tid);
mutex_unlock(&JFS_IP(ip)->commit_mutex);
mutex_unlock(&JFS_IP(dip)->commit_mutex);
goto out2;
}
/* update parent directory's link count corresponding
* to ".." entry of the target directory deleted
*/
dip->i_ctime = dip->i_mtime = CURRENT_TIME;
inode_dec_link_count(dip);
/*
* OS/2 could have created EA and/or ACL
*/
/* free EA from both persistent and working map */
if (JFS_IP(ip)->ea.flag & DXD_EXTENT) {
/* free EA pages */
txEA(tid, ip, &JFS_IP(ip)->ea, NULL);
}
JFS_IP(ip)->ea.flag = 0;
/* free ACL from both persistent and working map */
if (JFS_IP(ip)->acl.flag & DXD_EXTENT) {
/* free ACL pages */
txEA(tid, ip, &JFS_IP(ip)->acl, NULL);
}
JFS_IP(ip)->acl.flag = 0;
/* mark the target directory as deleted */
clear_nlink(ip);
mark_inode_dirty(ip);
rc = txCommit(tid, 2, &iplist[0], 0);
txEnd(tid);
mutex_unlock(&JFS_IP(ip)->commit_mutex);
mutex_unlock(&JFS_IP(dip)->commit_mutex);
/*
* Truncating the directory index table is not guaranteed. It
* may need to be done iteratively
*/
if (test_cflag(COMMIT_Stale, dip)) {
if (dip->i_size > 1)
jfs_truncate_nolock(dip, 0);
clear_cflag(COMMIT_Stale, dip);
}
out2:
free_UCSname(&dname);
out:
jfs_info("jfs_rmdir: rc:%d", rc);
return rc;
}
long jfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
{
struct inode *inode = filp->f_dentry->d_inode;
struct jfs_inode_info *jfs_inode = JFS_IP(inode);
unsigned int flags;
switch (cmd) {
case JFS_IOC_GETFLAGS:
jfs_get_inode_flags(jfs_inode);
flags = jfs_inode->mode2 & JFS_FL_USER_VISIBLE;
flags = jfs_map_ext2(flags, 0);
return put_user(flags, (int __user *) arg);
case JFS_IOC_SETFLAGS: {
unsigned int oldflags;
int err;
err = mnt_want_write(filp->f_path.mnt);
if (err)
return err;
if (!inode_owner_or_capable(inode)) {
err = -EACCES;
goto setflags_out;
}
if (get_user(flags, (int __user *) arg)) {
err = -EFAULT;
goto setflags_out;
}
flags = jfs_map_ext2(flags, 1);
if (!S_ISDIR(inode->i_mode))
flags &= ~JFS_DIRSYNC_FL;
/* Is it quota file? Do not allow user to mess with it */
if (IS_NOQUOTA(inode)) {
err = -EPERM;
goto setflags_out;
}
/* Lock against other parallel changes of flags */
mutex_lock(&inode->i_mutex);
jfs_get_inode_flags(jfs_inode);
oldflags = jfs_inode->mode2;
/*
* The IMMUTABLE and APPEND_ONLY flags can only be changed by
* the relevant capability.
*/
if ((oldflags & JFS_IMMUTABLE_FL) ||
((flags ^ oldflags) &
(JFS_APPEND_FL | JFS_IMMUTABLE_FL))) {
if (!capable(CAP_LINUX_IMMUTABLE)) {
mutex_unlock(&inode->i_mutex);
err = -EPERM;
goto setflags_out;
}
}
flags = flags & JFS_FL_USER_MODIFIABLE;
flags |= oldflags & ~JFS_FL_USER_MODIFIABLE;
jfs_inode->mode2 = flags;
jfs_set_inode_flags(inode);
mutex_unlock(&inode->i_mutex);
inode->i_ctime = CURRENT_TIME_SEC;
mark_inode_dirty(inode);
setflags_out:
mnt_drop_write_file(filp);
return err;
}
default:
return -ENOTTY;
}
}
/*
* NAME: jfs_unlink(dip, dentry)
*
* FUNCTION: remove a link to object <vp> named by <name>
* from parent directory <dvp>
*
* PARAMETER: dip - inode of parent directory
* dentry - dentry of object to be removed
*
* RETURN: errors from subroutines
*
* note:
* temporary file: if one or more processes have the file open
* when the last link is removed, the link will be removed before
* unlink() returns, but the removal of the file contents will be
* postponed until all references to the files are closed.
*
* JFS does NOT support unlink() on directories.
*
*/
static int jfs_unlink(struct inode *dip, struct dentry *dentry)
{
int rc;
tid_t tid; /* transaction id */
struct inode *ip = dentry->d_inode;
ino_t ino;
struct component_name dname; /* object name */
struct inode *iplist[2];
struct tblock *tblk;
s64 new_size = 0;
int commit_flag;
jfs_info("jfs_unlink: dip:0x%p name:%s", dip, dentry->d_name.name);
/* Init inode for quota operations. */
dquot_initialize(dip);
dquot_initialize(ip);
if ((rc = get_UCSname(&dname, dentry)))
goto out;
IWRITE_LOCK(ip, RDWRLOCK_NORMAL);
tid = txBegin(dip->i_sb, 0);
mutex_lock_nested(&JFS_IP(dip)->commit_mutex, COMMIT_MUTEX_PARENT);
mutex_lock_nested(&JFS_IP(ip)->commit_mutex, COMMIT_MUTEX_CHILD);
iplist[0] = dip;
iplist[1] = ip;
/*
* delete the entry of target file from parent directory
*/
ino = ip->i_ino;
if ((rc = dtDelete(tid, dip, &dname, &ino, JFS_REMOVE))) {
jfs_err("jfs_unlink: dtDelete returned %d", rc);
if (rc == -EIO)
txAbort(tid, 1); /* Marks FS Dirty */
txEnd(tid);
mutex_unlock(&JFS_IP(ip)->commit_mutex);
mutex_unlock(&JFS_IP(dip)->commit_mutex);
IWRITE_UNLOCK(ip);
goto out1;
}
ASSERT(ip->i_nlink);
ip->i_ctime = dip->i_ctime = dip->i_mtime = CURRENT_TIME;
mark_inode_dirty(dip);
/* update target's inode */
inode_dec_link_count(ip);
/*
* commit zero link count object
*/
if (ip->i_nlink == 0) {
assert(!test_cflag(COMMIT_Nolink, ip));
/* free block resources */
if ((new_size = commitZeroLink(tid, ip)) < 0) {
txAbort(tid, 1); /* Marks FS Dirty */
txEnd(tid);
mutex_unlock(&JFS_IP(ip)->commit_mutex);
mutex_unlock(&JFS_IP(dip)->commit_mutex);
IWRITE_UNLOCK(ip);
rc = new_size;
goto out1;
}
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_DELETE;
tblk->u.ip = ip;
}
/*
* Incomplete truncate of file data can
* result in timing problems unless we synchronously commit the
* transaction.
*/
if (new_size)
commit_flag = COMMIT_SYNC;
else
commit_flag = 0;
/*
* If xtTruncate was incomplete, commit synchronously to avoid
* timing complications
*/
rc = txCommit(tid, 2, &iplist[0], commit_flag);
txEnd(tid);
mutex_unlock(&JFS_IP(ip)->commit_mutex);
mutex_unlock(&JFS_IP(dip)->commit_mutex);
while (new_size && (rc == 0)) {
tid = txBegin(dip->i_sb, 0);
mutex_lock(&JFS_IP(ip)->commit_mutex);
new_size = xtTruncate_pmap(tid, ip, new_size);
if (new_size < 0) {
txAbort(tid, 1); /* Marks FS Dirty */
rc = new_size;
} else
rc = txCommit(tid, 2, &iplist[0], COMMIT_SYNC);
txEnd(tid);
mutex_unlock(&JFS_IP(ip)->commit_mutex);
}
if (ip->i_nlink == 0)
set_cflag(COMMIT_Nolink, ip);
IWRITE_UNLOCK(ip);
/*
* Truncating the directory index table is not guaranteed. It
* may need to be done iteratively
*/
if (test_cflag(COMMIT_Stale, dip)) {
if (dip->i_size > 1)
jfs_truncate_nolock(dip, 0);
clear_cflag(COMMIT_Stale, dip);
}
out1:
free_UCSname(&dname);
out:
jfs_info("jfs_unlink: rc:%d", rc);
return rc;
}
static int jfs_follow_link(struct dentry *dentry, struct nameidata *nd)
{
char *s = JFS_IP(dentry->d_inode)->i_inline;
return vfs_follow_link(nd, s);
}
/*
* NAME: jfs_create(dip, dentry, mode)
*
* FUNCTION: create a regular file in the parent directory <dip>
* with name = <from dentry> and mode = <mode>
*
* PARAMETER: dip - parent directory vnode
* dentry - dentry of new file
* mode - create mode (rwxrwxrwx).
* nd- nd struct
*
* RETURN: Errors from subroutines
*
*/
static int jfs_create(struct inode *dip, struct dentry *dentry, int mode,
struct nameidata *nd)
{
int rc = 0;
tid_t tid; /* transaction id */
struct inode *ip = NULL; /* child directory inode */
ino_t ino;
struct component_name dname; /* child directory name */
struct btstack btstack;
struct inode *iplist[2];
struct tblock *tblk;
jfs_info("jfs_create: dip:0x%p name:%s", dip, dentry->d_name.name);
dquot_initialize(dip);
/*
* search parent directory for entry/freespace
* (dtSearch() returns parent directory page pinned)
*/
if ((rc = get_UCSname(&dname, dentry)))
goto out1;
/*
* Either iAlloc() or txBegin() may block. Deadlock can occur if we
* block there while holding dtree page, so we allocate the inode &
* begin the transaction before we search the directory.
*/
ip = ialloc(dip, mode);
if (IS_ERR(ip)) {
rc = PTR_ERR(ip);
goto out2;
}
tid = txBegin(dip->i_sb, 0);
mutex_lock_nested(&JFS_IP(dip)->commit_mutex, COMMIT_MUTEX_PARENT);
mutex_lock_nested(&JFS_IP(ip)->commit_mutex, COMMIT_MUTEX_CHILD);
rc = jfs_init_acl(tid, ip, dip);
if (rc)
goto out3;
rc = jfs_init_security(tid, ip, dip, &dentry->d_name);
if (rc) {
txAbort(tid, 0);
goto out3;
}
if ((rc = dtSearch(dip, &dname, &ino, &btstack, JFS_CREATE))) {
jfs_err("jfs_create: dtSearch returned %d", rc);
txAbort(tid, 0);
goto out3;
}
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_CREATE;
tblk->ino = ip->i_ino;
tblk->u.ixpxd = JFS_IP(ip)->ixpxd;
iplist[0] = dip;
iplist[1] = ip;
/*
* initialize the child XAD tree root in-line in inode
*/
xtInitRoot(tid, ip);
/*
* create entry in parent directory for child directory
* (dtInsert() releases parent directory page)
*/
ino = ip->i_ino;
if ((rc = dtInsert(tid, dip, &dname, &ino, &btstack))) {
if (rc == -EIO) {
jfs_err("jfs_create: dtInsert returned -EIO");
txAbort(tid, 1); /* Marks Filesystem dirty */
} else
txAbort(tid, 0); /* Filesystem full */
goto out3;
}
ip->i_op = &jfs_file_inode_operations;
ip->i_fop = &jfs_file_operations;
ip->i_mapping->a_ops = &jfs_aops;
mark_inode_dirty(ip);
dip->i_ctime = dip->i_mtime = CURRENT_TIME;
mark_inode_dirty(dip);
rc = txCommit(tid, 2, &iplist[0], 0);
out3:
txEnd(tid);
mutex_unlock(&JFS_IP(ip)->commit_mutex);
mutex_unlock(&JFS_IP(dip)->commit_mutex);
if (rc) {
free_ea_wmap(ip);
ip->i_nlink = 0;
unlock_new_inode(ip);
iput(ip);
} else {
d_instantiate(dentry, ip);
unlock_new_inode(ip);
}
out2:
free_UCSname(&dname);
out1:
jfs_info("jfs_create: rc:%d", rc);
return rc;
}
/*
* NAME: jfs_mknod
*
* FUNCTION: Create a special file (device)
*/
int jfs_mknod(struct inode *dir, struct dentry *dentry, int mode, int rdev)
{
struct btstack btstack;
struct component_name dname;
ino_t ino;
struct inode *ip;
struct inode *iplist[2];
int rc;
tid_t tid;
struct tblock *tblk;
jFYI(1, ("jfs_mknod: %s\n", dentry->d_name.name));
if ((rc = get_UCSname(&dname, dentry, JFS_SBI(dir->i_sb)->nls_tab)))
goto out;
ip = ialloc(dir, mode);
if (ip == NULL) {
rc = ENOSPC;
goto out1;
}
tid = txBegin(dir->i_sb, 0);
down(&JFS_IP(dir)->commit_sem);
down(&JFS_IP(ip)->commit_sem);
if ((rc = dtSearch(dir, &dname, &ino, &btstack, JFS_CREATE)))
goto out3;
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_CREATE;
tblk->ip = ip;
ino = ip->i_ino;
if ((rc = dtInsert(tid, dir, &dname, &ino, &btstack)))
goto out3;
ip->i_op = &jfs_file_inode_operations;
init_special_inode(ip, ip->i_mode, rdev);
insert_inode_hash(ip);
mark_inode_dirty(ip);
d_instantiate(dentry, ip);
dir->i_ctime = dir->i_mtime = CURRENT_TIME;
mark_inode_dirty(dir);
iplist[0] = dir;
iplist[1] = ip;
rc = txCommit(tid, 2, iplist, 0);
out3:
txEnd(tid);
up(&JFS_IP(ip)->commit_sem);
up(&JFS_IP(dir)->commit_sem);
if (rc) {
ip->i_nlink = 0;
iput(ip);
}
out1:
free_UCSname(&dname);
out:
jFYI(1, ("jfs_mknod: returning %d\n", rc));
return -rc;
}
static int jfs_symlink(struct inode *dip, struct dentry *dentry,
const char *name)
{
int rc;
tid_t tid;
ino_t ino = 0;
struct component_name dname;
int ssize; /* source pathname size */
struct btstack btstack;
struct inode *ip = dentry->d_inode;
unchar *i_fastsymlink;
s64 xlen = 0;
int bmask = 0, xsize;
s64 extent = 0, xaddr;
struct metapage *mp;
struct super_block *sb;
struct tblock *tblk;
struct inode *iplist[2];
jfs_info("jfs_symlink: dip:0x%p name:%s", dip, name);
dquot_initialize(dip);
ssize = strlen(name) + 1;
/*
* search parent directory for entry/freespace
* (dtSearch() returns parent directory page pinned)
*/
if ((rc = get_UCSname(&dname, dentry)))
goto out1;
/*
* allocate on-disk/in-memory inode for symbolic link:
* (iAlloc() returns new, locked inode)
*/
ip = ialloc(dip, S_IFLNK | 0777);
if (IS_ERR(ip)) {
rc = PTR_ERR(ip);
goto out2;
}
tid = txBegin(dip->i_sb, 0);
mutex_lock_nested(&JFS_IP(dip)->commit_mutex, COMMIT_MUTEX_PARENT);
mutex_lock_nested(&JFS_IP(ip)->commit_mutex, COMMIT_MUTEX_CHILD);
rc = jfs_init_security(tid, ip, dip, &dentry->d_name);
if (rc)
goto out3;
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_CREATE;
tblk->ino = ip->i_ino;
tblk->u.ixpxd = JFS_IP(ip)->ixpxd;
/* fix symlink access permission
* (dir_create() ANDs in the u.u_cmask,
* but symlinks really need to be 777 access)
*/
ip->i_mode |= 0777;
/*
* write symbolic link target path name
*/
xtInitRoot(tid, ip);
/*
* write source path name inline in on-disk inode (fast symbolic link)
*/
if (ssize <= IDATASIZE) {
ip->i_op = &jfs_fast_symlink_inode_operations;
i_fastsymlink = JFS_IP(ip)->i_inline;
memcpy(i_fastsymlink, name, ssize);
ip->i_size = ssize - 1;
/*
* if symlink is > 128 bytes, we don't have the space to
* store inline extended attributes
*/
if (ssize > sizeof (JFS_IP(ip)->i_inline))
JFS_IP(ip)->mode2 &= ~INLINEEA;
jfs_info("jfs_symlink: fast symlink added ssize:%d name:%s ",
ssize, name);
}
/*
* write source path name in a single extent
*/
else {
jfs_info("jfs_symlink: allocate extent ip:0x%p", ip);
ip->i_op = &jfs_symlink_inode_operations;
ip->i_mapping->a_ops = &jfs_aops;
/*
* even though the data of symlink object (source
* path name) is treated as non-journaled user data,
* it is read/written thru buffer cache for performance.
*/
sb = ip->i_sb;
bmask = JFS_SBI(sb)->bsize - 1;
xsize = (ssize + bmask) & ~bmask;
xaddr = 0;
xlen = xsize >> JFS_SBI(sb)->l2bsize;
if ((rc = xtInsert(tid, ip, 0, 0, xlen, &xaddr, 0))) {
txAbort(tid, 0);
goto out3;
}
extent = xaddr;
ip->i_size = ssize - 1;
while (ssize) {
/* This is kind of silly since PATH_MAX == 4K */
int copy_size = min(ssize, PSIZE);
mp = get_metapage(ip, xaddr, PSIZE, 1);
if (mp == NULL) {
xtTruncate(tid, ip, 0, COMMIT_PWMAP);
rc = -EIO;
txAbort(tid, 0);
goto out3;
}
memcpy(mp->data, name, copy_size);
flush_metapage(mp);
ssize -= copy_size;
name += copy_size;
xaddr += JFS_SBI(sb)->nbperpage;
}
}
/*
* create entry for symbolic link in parent directory
*/
rc = dtSearch(dip, &dname, &ino, &btstack, JFS_CREATE);
if (rc == 0) {
ino = ip->i_ino;
rc = dtInsert(tid, dip, &dname, &ino, &btstack);
}
if (rc) {
if (xlen)
xtTruncate(tid, ip, 0, COMMIT_PWMAP);
txAbort(tid, 0);
/* discard new inode */
goto out3;
}
mark_inode_dirty(ip);
dip->i_ctime = dip->i_mtime = CURRENT_TIME;
mark_inode_dirty(dip);
/*
* commit update of parent directory and link object
*/
iplist[0] = dip;
iplist[1] = ip;
rc = txCommit(tid, 2, &iplist[0], 0);
out3:
txEnd(tid);
mutex_unlock(&JFS_IP(ip)->commit_mutex);
mutex_unlock(&JFS_IP(dip)->commit_mutex);
if (rc) {
free_ea_wmap(ip);
ip->i_nlink = 0;
unlock_new_inode(ip);
iput(ip);
} else {
d_instantiate(dentry, ip);
unlock_new_inode(ip);
}
out2:
free_UCSname(&dname);
out1:
jfs_info("jfs_symlink: rc:%d", rc);
return rc;
}
long jfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
{
struct inode *inode = file_inode(filp);
struct jfs_inode_info *jfs_inode = JFS_IP(inode);
unsigned int flags;
switch (cmd) {
case JFS_IOC_GETFLAGS:
flags = jfs_inode->mode2 & JFS_FL_USER_VISIBLE;
flags = jfs_map_ext2(flags, 0);
return put_user(flags, (int __user *) arg);
case JFS_IOC_SETFLAGS: {
unsigned int oldflags;
int err;
err = mnt_want_write_file(filp);
if (err)
return err;
if (!inode_owner_or_capable(inode)) {
err = -EACCES;
goto setflags_out;
}
if (get_user(flags, (int __user *) arg)) {
err = -EFAULT;
goto setflags_out;
}
flags = jfs_map_ext2(flags, 1);
if (!S_ISDIR(inode->i_mode))
flags &= ~JFS_DIRSYNC_FL;
/* Is it quota file? Do not allow user to mess with it */
if (IS_NOQUOTA(inode)) {
err = -EPERM;
goto setflags_out;
}
/* Lock against other parallel changes of flags */
inode_lock(inode);
oldflags = jfs_inode->mode2;
/*
* The IMMUTABLE and APPEND_ONLY flags can only be changed by
* the relevant capability.
*/
if ((oldflags & JFS_IMMUTABLE_FL) ||
((flags ^ oldflags) &
(JFS_APPEND_FL | JFS_IMMUTABLE_FL))) {
if (!capable(CAP_LINUX_IMMUTABLE)) {
inode_unlock(inode);
err = -EPERM;
goto setflags_out;
}
}
flags = flags & JFS_FL_USER_MODIFIABLE;
flags |= oldflags & ~JFS_FL_USER_MODIFIABLE;
jfs_inode->mode2 = flags;
jfs_set_inode_flags(inode);
inode_unlock(inode);
inode->i_ctime = current_time(inode);
mark_inode_dirty(inode);
setflags_out:
mnt_drop_write_file(filp);
return err;
}
case FITRIM:
{
struct super_block *sb = inode->i_sb;
struct request_queue *q = bdev_get_queue(sb->s_bdev);
struct fstrim_range range;
s64 ret = 0;
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
if (!blk_queue_discard(q)) {
jfs_warn("FITRIM not supported on device");
return -EOPNOTSUPP;
}
if (copy_from_user(&range, (struct fstrim_range __user *)arg,
sizeof(range)))
return -EFAULT;
range.minlen = max_t(unsigned int, range.minlen,
q->limits.discard_granularity);
ret = jfs_ioc_trim(inode, &range);
if (ret < 0)
return ret;
if (copy_to_user((struct fstrim_range __user *)arg, &range,
sizeof(range)))
return -EFAULT;
return 0;
}
default:
return -ENOTTY;
}
}
/*
* NAME: ialloc()
*
* FUNCTION: Allocate a new inode
*
*/
struct inode *ialloc(struct inode *parent, umode_t mode)
{
struct super_block *sb = parent->i_sb;
struct inode *inode;
struct jfs_inode_info *jfs_inode;
int rc;
inode = new_inode(sb);
if (!inode) {
jfs_warn("ialloc: new_inode returned NULL!");
return ERR_PTR(-ENOMEM);
}
jfs_inode = JFS_IP(inode);
rc = diAlloc(parent, S_ISDIR(mode), inode);
if (rc) {
jfs_warn("ialloc: diAlloc returned %d!", rc);
if (rc == -EIO)
make_bad_inode(inode);
iput(inode);
return ERR_PTR(rc);
}
inode->i_uid = current_fsuid();
if (parent->i_mode & S_ISGID) {
inode->i_gid = parent->i_gid;
if (S_ISDIR(mode))
mode |= S_ISGID;
} else
inode->i_gid = current_fsgid();
/*
* New inodes need to save sane values on disk when
* uid & gid mount options are used
*/
jfs_inode->saved_uid = inode->i_uid;
jfs_inode->saved_gid = inode->i_gid;
/*
* Allocate inode to quota.
*/
if (DQUOT_ALLOC_INODE(inode)) {
DQUOT_DROP(inode);
inode->i_flags |= S_NOQUOTA;
inode->i_nlink = 0;
iput(inode);
return ERR_PTR(-EDQUOT);
}
inode->i_mode = mode;
/* inherit flags from parent */
jfs_inode->mode2 = JFS_IP(parent)->mode2 & JFS_FL_INHERIT;
if (S_ISDIR(mode)) {
jfs_inode->mode2 |= IDIRECTORY;
jfs_inode->mode2 &= ~JFS_DIRSYNC_FL;
}
else {
jfs_inode->mode2 |= INLINEEA | ISPARSE;
if (S_ISLNK(mode))
jfs_inode->mode2 &= ~(JFS_IMMUTABLE_FL|JFS_APPEND_FL);
}
jfs_inode->mode2 |= mode;
inode->i_blocks = 0;
inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
jfs_inode->otime = inode->i_ctime.tv_sec;
inode->i_generation = JFS_SBI(sb)->gengen++;
jfs_inode->cflag = 0;
/* Zero remaining fields */
memset(&jfs_inode->acl, 0, sizeof(dxd_t));
memset(&jfs_inode->ea, 0, sizeof(dxd_t));
jfs_inode->next_index = 0;
jfs_inode->acltype = 0;
jfs_inode->btorder = 0;
jfs_inode->btindex = 0;
jfs_inode->bxflag = 0;
jfs_inode->blid = 0;
jfs_inode->atlhead = 0;
jfs_inode->atltail = 0;
jfs_inode->xtlid = 0;
jfs_set_inode_flags(inode);
jfs_info("ialloc returns inode = 0x%p\n", inode);
return inode;
}
static int
jfs_get_blocks(struct inode *ip, sector_t lblock, unsigned long max_blocks,
struct buffer_head *bh_result, int create)
{
s64 lblock64 = lblock;
int no_size_check = 0;
int rc = 0;
int take_locks;
xad_t xad;
s64 xaddr;
int xflag;
s32 xlen;
/*
* If this is a special inode (imap, dmap) or directory,
* the lock should already be taken
*/
take_locks = ((JFS_IP(ip)->fileset != AGGREGATE_I) &&
!S_ISDIR(ip->i_mode));
/*
* Take appropriate lock on inode
*/
if (take_locks) {
if (create)
IWRITE_LOCK(ip);
else
IREAD_LOCK(ip);
}
/*
* A directory's "data" is the inode index table, but i_size is the
* size of the d-tree, so don't check the offset against i_size
*/
if (S_ISDIR(ip->i_mode))
no_size_check = 1;
if ((no_size_check ||
((lblock64 << ip->i_sb->s_blocksize_bits) < ip->i_size)) &&
(xtLookup(ip, lblock64, max_blocks, &xflag, &xaddr, &xlen, no_size_check)
== 0) && xlen) {
if (xflag & XAD_NOTRECORDED) {
if (!create)
/*
* Allocated but not recorded, read treats
* this as a hole
*/
goto unlock;
#ifdef _JFS_4K
XADoffset(&xad, lblock64);
XADlength(&xad, xlen);
XADaddress(&xad, xaddr);
#else /* _JFS_4K */
/*
* As long as block size = 4K, this isn't a problem.
* We should mark the whole page not ABNR, but how
* will we know to mark the other blocks BH_New?
*/
BUG();
#endif /* _JFS_4K */
rc = extRecord(ip, &xad);
if (rc)
goto unlock;
set_buffer_new(bh_result);
}
map_bh(bh_result, ip->i_sb, xaddr);
bh_result->b_size = xlen << ip->i_blkbits;
goto unlock;
}
if (!create)
goto unlock;
/*
* Allocate a new block
*/
#ifdef _JFS_4K
if ((rc = extHint(ip, lblock64 << ip->i_sb->s_blocksize_bits, &xad)))
goto unlock;
rc = extAlloc(ip, max_blocks, lblock64, &xad, FALSE);
if (rc)
goto unlock;
set_buffer_new(bh_result);
map_bh(bh_result, ip->i_sb, addressXAD(&xad));
bh_result->b_size = lengthXAD(&xad) << ip->i_blkbits;
#else /* _JFS_4K */
/*
* We need to do whatever it takes to keep all but the last buffers
* in 4K pages - see jfs_write.c
*/
BUG();
#endif /* _JFS_4K */
unlock:
/*
* Release lock on inode
*/
if (take_locks) {
if (create)
IWRITE_UNLOCK(ip);
else
IREAD_UNLOCK(ip);
}
return rc;
}
static int jfs_readlink(struct dentry *dentry, char *buffer, int buflen)
{
char *s = JFS_IP(dentry->d_inode)->i_inline;
return vfs_readlink(dentry, buffer, buflen, s);
}
/*
* NAME: jfs_create(dip, dentry, mode)
*
* FUNCTION: create a regular file in the parent directory <dip>
* with name = <from dentry> and mode = <mode>
*
* PARAMETER: dip - parent directory vnode
* dentry - dentry of new file
* mode - create mode (rwxrwxrwx).
*
* RETURN: Errors from subroutines
*
*/
int jfs_create(struct inode *dip, struct dentry *dentry, int mode)
{
int rc = 0;
tid_t tid; /* transaction id */
struct inode *ip = NULL; /* child directory inode */
ino_t ino;
struct component_name dname; /* child directory name */
struct btstack btstack;
struct inode *iplist[2];
struct tblock *tblk;
jFYI(1, ("jfs_create: dip:0x%p name:%s\n", dip, dentry->d_name.name));
/*
* search parent directory for entry/freespace
* (dtSearch() returns parent directory page pinned)
*/
if ((rc = get_UCSname(&dname, dentry, JFS_SBI(dip->i_sb)->nls_tab)))
goto out1;
/*
* Either iAlloc() or txBegin() may block. Deadlock can occur if we
* block there while holding dtree page, so we allocate the inode &
* begin the transaction before we search the directory.
*/
ip = ialloc(dip, mode);
if (ip == NULL) {
rc = ENOSPC;
goto out2;
}
tid = txBegin(dip->i_sb, 0);
down(&JFS_IP(dip)->commit_sem);
down(&JFS_IP(ip)->commit_sem);
if ((rc = dtSearch(dip, &dname, &ino, &btstack, JFS_CREATE))) {
jERROR(1, ("jfs_create: dtSearch returned %d\n", rc));
goto out3;
}
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_CREATE;
tblk->ip = ip;
iplist[0] = dip;
iplist[1] = ip;
/*
* initialize the child XAD tree root in-line in inode
*/
xtInitRoot(tid, ip);
/*
* create entry in parent directory for child directory
* (dtInsert() releases parent directory page)
*/
ino = ip->i_ino;
if ((rc = dtInsert(tid, dip, &dname, &ino, &btstack))) {
jERROR(1, ("jfs_create: dtInsert returned %d\n", rc));
if (rc == EIO)
txAbort(tid, 1); /* Marks Filesystem dirty */
else
txAbort(tid, 0); /* Filesystem full */
goto out3;
}
ip->i_op = &jfs_file_inode_operations;
ip->i_fop = &jfs_file_operations;
ip->i_mapping->a_ops = &jfs_aops;
insert_inode_hash(ip);
mark_inode_dirty(ip);
d_instantiate(dentry, ip);
dip->i_ctime = dip->i_mtime = CURRENT_TIME;
mark_inode_dirty(dip);
rc = txCommit(tid, 2, &iplist[0], 0);
out3:
txEnd(tid);
up(&JFS_IP(dip)->commit_sem);
up(&JFS_IP(ip)->commit_sem);
if (rc) {
ip->i_nlink = 0;
iput(ip);
}
out2:
free_UCSname(&dname);
out1:
jFYI(1, ("jfs_create: rc:%d\n", -rc));
return -rc;
}