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));
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);
}
/*
* NAME: jfs_strtoUCS()
*
* FUNCTION: Convert character string to unicode string
*
*/
static int jfs_strtoUCS(wchar_t * to, const unsigned char *from, int len,
struct nls_table *codepage)
{
int charlen;
int i;
if (codepage) {
for (i = 0; len && *from; i++, from += charlen, len -= charlen)
{
charlen = codepage->char2uni(from, len, &to[i]);
if (charlen < 1) {
jfs_err("jfs_strtoUCS: char2uni returned %d.",
charlen);
jfs_err("charset = %s, char = 0x%x",
codepage->charset, *from);
return charlen;
}
}
} else {
for (i = 0; (i < len) && from[i]; i++)
to[i] = (wchar_t) from[i];
}
to[i] = 0;
return i;
}
static struct dentry *jfs_lookup(struct inode *dip, struct dentry *dentry, struct nameidata *nd)
{
struct btstack btstack;
ino_t inum;
struct inode *ip;
struct component_name key;
int rc;
jfs_info("jfs_lookup: name = %s", dentry->d_name.name);
if ((rc = get_UCSname(&key, dentry)))
return ERR_PTR(rc);
rc = dtSearch(dip, &key, &inum, &btstack, JFS_LOOKUP);
free_UCSname(&key);
if (rc == -ENOENT) {
ip = NULL;
} else if (rc) {
jfs_err("jfs_lookup: dtSearch returned %d", rc);
ip = ERR_PTR(rc);
} else {
ip = jfs_iget(dip->i_sb, inum);
if (IS_ERR(ip))
jfs_err("jfs_lookup: iget failed on inum %d", (uint)inum);
}
return d_splice_alias(ip, dentry);
}
/*
* NAME: jfs_mount_rw(sb, remount)
*
* FUNCTION: Completes read-write mount, or remounts read-only volume
* as read-write
*/
int jfs_mount_rw(struct super_block *sb, int remount)
{
struct jfs_sb_info *sbi = JFS_SBI(sb);
int rc;
/*
* If we are re-mounting a previously read-only volume, we want to
* re-read the inode and block maps, since fsck.jfs may have updated
* them.
*/
if (remount) {
if (chkSuper(sb) || (sbi->state != FM_CLEAN))
return -EINVAL;
truncate_inode_pages(sbi->ipimap->i_mapping, 0);
truncate_inode_pages(sbi->ipbmap->i_mapping, 0);
diUnmount(sbi->ipimap, 1);
if ((rc = diMount(sbi->ipimap))) {
jfs_err("jfs_mount_rw: diMount failed!");
return rc;
}
dbUnmount(sbi->ipbmap, 1);
if ((rc = dbMount(sbi->ipbmap))) {
jfs_err("jfs_mount_rw: dbMount failed!");
return rc;
}
}
/*
* open/initialize log
*/
if ((rc = lmLogOpen(sb)))
return rc;
/*
* update file system superblock;
*/
if ((rc = updateSuper(sb, FM_MOUNT))) {
jfs_err("jfs_mount: updateSuper failed w/rc = %d", rc);
lmLogClose(sb);
return rc;
}
/*
* write MOUNT log record of the file system
*/
logMOUNT(sb);
/*
* Set page cache allocation policy
*/
mapping_set_gfp_mask(sb->s_bdev->bd_inode->i_mapping, GFP_NOFS);
return rc;
}
static struct dentry *jfs_lookup(struct inode *dip, struct dentry *dentry, struct nameidata *nd)
{
struct btstack btstack;
ino_t inum;
struct inode *ip;
struct component_name key;
const char *name = dentry->d_name.name;
int len = dentry->d_name.len;
int rc;
jfs_info("jfs_lookup: name = %s", name);
if ((name[0] == '.') && (len == 1))
inum = dip->i_ino;
else if (strcmp(name, "..") == 0)
inum = PARENT(dip);
else {
if ((rc = get_UCSname(&key, dentry)))
return ERR_PTR(rc);
rc = dtSearch(dip, &key, &inum, &btstack, JFS_LOOKUP);
free_UCSname(&key);
if (rc == -ENOENT) {
d_add(dentry, NULL);
return ERR_PTR(0);
} else if (rc) {
jfs_err("jfs_lookup: dtSearch returned %d", rc);
return ERR_PTR(rc);
}
}
ip = iget(dip->i_sb, inum);
if (ip == NULL || is_bad_inode(ip)) {
jfs_err("jfs_lookup: iget failed on inum %d", (uint) inum);
if (ip)
iput(ip);
return ERR_PTR(-EACCES);
}
if (JFS_SBI(dip->i_sb)->mntflag & JFS_OS2)
dentry->d_op = &jfs_ci_dentry_operations;
dentry = d_splice_alias(ip, dentry);
if (dentry && (JFS_SBI(dip->i_sb)->mntflag & JFS_OS2))
dentry->d_op = &jfs_ci_dentry_operations;
return dentry;
}
static struct dentry *jfs_lookup(struct inode *dip, struct dentry *dentry)
{
struct btstack btstack;
ino_t inum;
struct inode *ip;
struct component_name key;
const char *name = dentry->d_name.name;
int len = dentry->d_name.len;
int rc;
jfs_info("jfs_lookup: name = %s", name);
if ((name[0] == '.') && (len == 1))
inum = dip->i_ino;
else if (strcmp(name, "..") == 0)
inum = PARENT(dip);
else {
if ((rc =
get_UCSname(&key, dentry, JFS_SBI(dip->i_sb)->nls_tab)))
return ERR_PTR(rc);
rc = dtSearch(dip, &key, &inum, &btstack, JFS_LOOKUP);
free_UCSname(&key);
if (rc == -ENOENT) {
d_add(dentry, NULL);
return ERR_PTR(0);
} else if (rc) {
jfs_err("jfs_lookup: dtSearch returned %d", rc);
return ERR_PTR(rc);
}
}
ip = iget(dip->i_sb, inum);
if (ip == NULL) {
jfs_err("jfs_lookup: iget failed on inum %d", (uint) inum);
return ERR_PTR(-EACCES);
}
if (is_bad_inode(ip)) {
jfs_err("jfs_lookup: iget returned bad inode, inum = %d",
(uint) inum);
iput(ip);
return ERR_PTR(-EACCES);
}
d_add(dentry, ip);
return ERR_PTR(0);
}
void jfs_dirty_inode(struct inode *inode)
{
static int noisy = 5;
if (isReadOnly(inode)) {
if (!special_file(inode->i_mode) && noisy) {
/* kernel allows writes to devices on read-only
* partitions and may try to mark inode dirty
*/
jfs_err("jfs_dirty_inode called on read-only volume");
jfs_err("Is remount racy?");
noisy--;
}
return;
}
set_cflag(COMMIT_Dirty, inode);
}
/*
* NAME: jfs_strtoUCS()
*
* FUNCTION: Convert character string to unicode string
*
*/
int jfs_strtoUCS(wchar_t * to,
const char *from, int len, struct nls_table *codepage)
{
int charlen;
int i;
for (i = 0; len && *from; i++, from += charlen, len -= charlen) {
charlen = codepage->char2uni(from, len, &to[i]);
if (charlen < 1) {
jfs_err("jfs_strtoUCS: char2uni returned %d.", charlen);
jfs_err("charset = %s, char = 0x%x",
codepage->charset, (unsigned char) *from);
return charlen;
}
}
to[i] = 0;
return i;
}
/*
* Workhorse of both fsync & write_inode
*/
int jfs_commit_inode(struct inode *inode, int wait)
{
int rc = 0;
tid_t tid;
static int noisy = 5;
jfs_info("In jfs_commit_inode, inode = 0x%p", inode);
/*
* Don't commit if inode has been committed since last being
* marked dirty, or if it has been deleted.
*/
if (inode->i_nlink == 0 || !test_cflag(COMMIT_Dirty, inode))
return 0;
if (isReadOnly(inode)) {
/* kernel allows writes to devices on read-only
* partitions and may think inode is dirty
*/
if (!special_file(inode->i_mode) && noisy) {
jfs_err("jfs_commit_inode(0x%p) called on "
"read-only volume", inode);
jfs_err("Is remount racy?");
noisy--;
}
return 0;
}
tid = txBegin(inode->i_sb, COMMIT_INODE);
mutex_lock(&JFS_IP(inode)->commit_mutex);
/*
* Retest inode state after taking commit_mutex
*/
if (inode->i_nlink && test_cflag(COMMIT_Dirty, inode))
rc = txCommit(tid, 1, &inode, wait ? COMMIT_SYNC : 0);
txEnd(tid);
mutex_unlock(&JFS_IP(inode)->commit_mutex);
return rc;
}
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);
}
static struct dentry *jfs_lookup(struct inode *dip, struct dentry *dentry, struct nameidata *nd)
{
struct btstack btstack;
ino_t inum;
struct inode *ip;
struct component_name key;
const char *name = dentry->d_name.name;
int len = dentry->d_name.len;
int rc;
jfs_info("jfs_lookup: name = %s", name);
if ((name[0] == '.') && (len == 1))
inum = dip->i_ino;
else if (strcmp(name, "..") == 0)
inum = PARENT(dip);
else {
if ((rc = get_UCSname(&key, dentry)))
return ERR_PTR(rc);
rc = dtSearch(dip, &key, &inum, &btstack, JFS_LOOKUP);
free_UCSname(&key);
if (rc == -ENOENT) {
d_add(dentry, NULL);
return NULL;
} else if (rc) {
jfs_err("jfs_lookup: dtSearch returned %d", rc);
return ERR_PTR(rc);
}
}
ip = jfs_iget(dip->i_sb, inum);
if (IS_ERR(ip)) {
jfs_err("jfs_lookup: iget failed on inum %d", (uint) inum);
return ERR_CAST(ip);
}
return d_splice_alias(ip, dentry);
}
int jfs_mount_rw(struct super_block *sb, int remount)
{
struct jfs_sb_info *sbi = JFS_SBI(sb);
int rc;
if (remount) {
if (chkSuper(sb) || (sbi->state != FM_CLEAN))
return -EINVAL;
truncate_inode_pages(sbi->ipimap->i_mapping, 0);
truncate_inode_pages(sbi->ipbmap->i_mapping, 0);
diUnmount(sbi->ipimap, 1);
if ((rc = diMount(sbi->ipimap))) {
jfs_err("jfs_mount_rw: diMount failed!");
return rc;
}
dbUnmount(sbi->ipbmap, 1);
if ((rc = dbMount(sbi->ipbmap))) {
jfs_err("jfs_mount_rw: dbMount failed!");
return rc;
}
}
if ((rc = lmLogOpen(sb)))
return rc;
if ((rc = updateSuper(sb, FM_MOUNT))) {
jfs_err("jfs_mount: updateSuper failed w/rc = %d", rc);
lmLogClose(sb);
return rc;
}
logMOUNT(sb);
return rc;
}
/*
* dbSync()
*/
int dbSync(struct inode *ipbmap)
{
struct dbmap_disk *dbmp_le;
struct bmap *bmp = JFS_SBI(ipbmap->i_sb)->bmap;
struct metapage *mp;
int i;
/*
* write bmap global control page
*/
/* get the buffer for the on-disk bmap descriptor. */
mp = read_metapage(ipbmap,
BMAPBLKNO << JFS_SBI(ipbmap->i_sb)->l2nbperpage,
PSIZE, 0);
if (mp == NULL) {
jfs_err("dbSync: read_metapage failed!");
return -EIO;
}
/* copy the in-memory version of the bmap to the on-disk version */
dbmp_le = (struct dbmap_disk *) mp->data;
dbmp_le->dn_mapsize = cpu_to_le64(bmp->db_mapsize);
dbmp_le->dn_nfree = cpu_to_le64(bmp->db_nfree);
dbmp_le->dn_l2nbperpage = cpu_to_le32(bmp->db_l2nbperpage);
dbmp_le->dn_numag = cpu_to_le32(bmp->db_numag);
dbmp_le->dn_maxlevel = cpu_to_le32(bmp->db_maxlevel);
dbmp_le->dn_maxag = cpu_to_le32(bmp->db_maxag);
dbmp_le->dn_agpref = cpu_to_le32(bmp->db_agpref);
dbmp_le->dn_aglevel = cpu_to_le32(bmp->db_aglevel);
dbmp_le->dn_agheight = cpu_to_le32(bmp->db_agheight);
dbmp_le->dn_agwidth = cpu_to_le32(bmp->db_agwidth);
dbmp_le->dn_agstart = cpu_to_le32(bmp->db_agstart);
dbmp_le->dn_agl2size = cpu_to_le32(bmp->db_agl2size);
for (i = 0; i < MAXAG; i++)
dbmp_le->dn_agfree[i] = cpu_to_le64(bmp->db_agfree[i]);
dbmp_le->dn_agsize = cpu_to_le64(bmp->db_agsize);
dbmp_le->dn_maxfreebud = bmp->db_maxfreebud;
/* write the buffer */
write_metapage(mp);
/*
* write out dirty pages of bmap
*/
filemap_write_and_wait(ipbmap->i_mapping);
diWriteSpecial(ipbmap, 0);
return (0);
}
/*
* updateSuper()
*
* update synchronously superblock if it is mounted read-write.
*/
int updateSuper(struct super_block *sb, uint state)
{
struct jfs_superblock *j_sb;
struct jfs_sb_info *sbi = JFS_SBI(sb);
struct buffer_head *bh;
int rc;
if (sbi->flag & JFS_NOINTEGRITY) {
if (state == FM_DIRTY) {
sbi->p_state = state;
return 0;
} else if (state == FM_MOUNT) {
sbi->p_state = sbi->state;
state = FM_DIRTY;
} else if (state == FM_CLEAN) {
state = sbi->p_state;
} else
jfs_err("updateSuper: bad state");
} else if (sbi->state == FM_DIRTY)
return 0;
if ((rc = readSuper(sb, &bh)))
return rc;
j_sb = (struct jfs_superblock *)bh->b_data;
j_sb->s_state = cpu_to_le32(state);
sbi->state = state;
if (state == FM_MOUNT) {
/* record log's dev_t and mount serial number */
j_sb->s_logdev = cpu_to_le32(sbi->log->bdev->bd_dev);
j_sb->s_logserial = cpu_to_le32(sbi->log->serial);
} else if (state == FM_CLEAN) {
/*
* If this volume is shared with OS/2, OS/2 will need to
* recalculate DASD usage, since we don't deal with it.
*/
if (j_sb->s_flag & cpu_to_le32(JFS_DASD_ENABLED))
j_sb->s_flag |= cpu_to_le32(JFS_DASD_PRIME);
}
mark_buffer_dirty(bh);
ll_rw_block(WRITE, 1, &bh);
wait_on_buffer(bh);
brelse(bh);
return 0;
}
void jfs_write_inode(struct inode *inode, int wait)
{
/*
* If COMMIT_DIRTY is not set, the inode isn't really dirty.
* It has been committed since the last change, but was still
* on the dirty inode list
*/
if (test_cflag(COMMIT_Nolink, inode) ||
!test_cflag(COMMIT_Dirty, inode))
return;
if (jfs_commit_inode(inode, wait)) {
jfs_err("jfs_write_inode: jfs_commit_inode failed!");
}
}
int updateSuper(struct super_block *sb, uint state)
{
struct jfs_superblock *j_sb;
struct jfs_sb_info *sbi = JFS_SBI(sb);
struct buffer_head *bh;
int rc;
if (sbi->flag & JFS_NOINTEGRITY) {
if (state == FM_DIRTY) {
sbi->p_state = state;
return 0;
} else if (state == FM_MOUNT) {
sbi->p_state = sbi->state;
state = FM_DIRTY;
} else if (state == FM_CLEAN) {
state = sbi->p_state;
} else
jfs_err("updateSuper: bad state");
} else if (sbi->state == FM_DIRTY)
return 0;
if ((rc = readSuper(sb, &bh)))
return rc;
j_sb = (struct jfs_superblock *)bh->b_data;
j_sb->s_state = cpu_to_le32(state);
sbi->state = state;
if (state == FM_MOUNT) {
j_sb->s_logdev = cpu_to_le32(new_encode_dev(sbi->log->bdev->bd_dev));
j_sb->s_logserial = cpu_to_le32(sbi->log->serial);
} else if (state == FM_CLEAN) {
if (j_sb->s_flag & cpu_to_le32(JFS_DASD_ENABLED))
j_sb->s_flag |= cpu_to_le32(JFS_DASD_PRIME);
}
mark_buffer_dirty(bh);
sync_dirty_buffer(bh);
brelse(bh);
return 0;
}
void jfs_write_inode(struct inode *inode, int wait)
{
if (test_cflag(COMMIT_Nolink, inode))
return;
/*
* If COMMIT_DIRTY is not set, the inode isn't really dirty.
* It has been committed since the last change, but was still
* on the dirty inode list.
*/
if (!test_cflag(COMMIT_Dirty, inode)) {
/* Make sure committed changes hit the disk */
jfs_flush_journal(JFS_SBI(inode->i_sb)->log, wait);
return;
}
if (jfs_commit_inode(inode, wait)) {
jfs_err("jfs_write_inode: jfs_commit_inode failed!");
}
}
int jfs_write_inode(struct inode *inode, struct writeback_control *wbc)
{
int wait = 1; /* XXX fix fsync and use wbc->sync_mode == WB_SYNC_ALL; */
if (test_cflag(COMMIT_Nolink, inode))
return 0;
/*
* If COMMIT_DIRTY is not set, the inode isn't really dirty.
* It has been committed since the last change, but was still
* on the dirty inode list.
*/
if (!test_cflag(COMMIT_Dirty, inode)) {
/* Make sure committed changes hit the disk */
jfs_flush_journal(JFS_SBI(inode->i_sb)->log, wait);
return 0;
}
if (jfs_commit_inode(inode, wait)) {
jfs_err("jfs_write_inode: jfs_commit_inode failed!");
return -EIO;
} else
return 0;
}
/*
* 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;
}
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 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);
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))) {
jfs_err("jfs_symlink: dtInsert returned %d", 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;
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 = &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:
jfs_info("jfs_symlink: rc:%d", rc);
return rc;
}
/*
* NAME: jfs_mount(sb)
*
* FUNCTION: vfs_mount()
*
* PARAMETER: sb - super block
*
* RETURN: -EBUSY - device already mounted or open for write
* -EBUSY - cvrdvp already mounted;
* -EBUSY - mount table full
* -ENOTDIR- cvrdvp not directory on a device mount
* -ENXIO - device open failure
*/
int jfs_mount(struct super_block *sb)
{
int rc = 0; /* Return code */
struct jfs_sb_info *sbi = JFS_SBI(sb);
struct inode *ipaimap = NULL;
struct inode *ipaimap2 = NULL;
struct inode *ipimap = NULL;
struct inode *ipbmap = NULL;
/*
* read/validate superblock
* (initialize mount inode from the superblock)
*/
if ((rc = chkSuper(sb))) {
goto errout20;
}
ipaimap = diReadSpecial(sb, AGGREGATE_I, 0);
if (ipaimap == NULL) {
jfs_err("jfs_mount: Faild to read AGGREGATE_I");
rc = -EIO;
goto errout20;
}
sbi->ipaimap = ipaimap;
jfs_info("jfs_mount: ipaimap:0x%p", ipaimap);
/*
* initialize aggregate inode allocation map
*/
if ((rc = diMount(ipaimap))) {
jfs_err("jfs_mount: diMount(ipaimap) failed w/rc = %d", rc);
goto errout21;
}
/*
* open aggregate block allocation map
*/
ipbmap = diReadSpecial(sb, BMAP_I, 0);
if (ipbmap == NULL) {
rc = -EIO;
goto errout22;
}
jfs_info("jfs_mount: ipbmap:0x%p", ipbmap);
sbi->ipbmap = ipbmap;
/*
* initialize aggregate block allocation map
*/
if ((rc = dbMount(ipbmap))) {
jfs_err("jfs_mount: dbMount failed w/rc = %d", rc);
goto errout22;
}
/*
* open the secondary aggregate inode allocation map
*
* This is a duplicate of the aggregate inode allocation map.
*
* hand craft a vfs in the same fashion as we did to read ipaimap.
* By adding INOSPEREXT (32) to the inode number, we are telling
* diReadSpecial that we are reading from the secondary aggregate
* inode table. This also creates a unique entry in the inode hash
* table.
*/
if ((sbi->mntflag & JFS_BAD_SAIT) == 0) {
ipaimap2 = diReadSpecial(sb, AGGREGATE_I, 1);
if (ipaimap2 == 0) {
jfs_err("jfs_mount: Faild to read AGGREGATE_I");
rc = -EIO;
goto errout35;
}
sbi->ipaimap2 = ipaimap2;
jfs_info("jfs_mount: ipaimap2:0x%p", ipaimap2);
/*
* initialize secondary aggregate inode allocation map
*/
if ((rc = diMount(ipaimap2))) {
jfs_err("jfs_mount: diMount(ipaimap2) failed, rc = %d",
rc);
goto errout35;
}
} else
/* Secondary aggregate inode table is not valid */
sbi->ipaimap2 = NULL;
/*
* mount (the only/single) fileset
*/
/*
* open fileset inode allocation map (aka fileset inode)
*/
ipimap = diReadSpecial(sb, FILESYSTEM_I, 0);
if (ipimap == NULL) {
jfs_err("jfs_mount: Failed to read FILESYSTEM_I");
/* open fileset secondary inode allocation map */
rc = -EIO;
goto errout40;
}
jfs_info("jfs_mount: ipimap:0x%p", ipimap);
/* map further access of per fileset inodes by the fileset inode */
sbi->ipimap = ipimap;
/* initialize fileset inode allocation map */
if ((rc = diMount(ipimap))) {
jfs_err("jfs_mount: diMount failed w/rc = %d", rc);
goto errout41;
}
goto out;
/*
* unwind on error
*/
errout41: /* close fileset inode allocation map inode */
diFreeSpecial(ipimap);
errout40: /* fileset closed */
/* close secondary aggregate inode allocation map */
if (ipaimap2) {
diUnmount(ipaimap2, 1);
diFreeSpecial(ipaimap2);
}
errout35:
/* close aggregate block allocation map */
dbUnmount(ipbmap, 1);
diFreeSpecial(ipbmap);
errout22: /* close aggregate inode allocation map */
diUnmount(ipaimap, 1);
errout21: /* close aggregate inodes */
diFreeSpecial(ipaimap);
errout20: /* aggregate closed */
out:
if (rc)
jfs_err("Mount JFS Failure: %d", rc);
return rc;
}
/*
* 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;
}
/*
* chkSuper()
*
* validate the superblock of the file system to be mounted and
* get the file system parameters.
*
* returns
* 0 with fragsize set if check successful
* error code if not successful
*/
static int chkSuper(struct super_block *sb)
{
int rc = 0;
struct jfs_sb_info *sbi = JFS_SBI(sb);
struct jfs_superblock *j_sb;
struct buffer_head *bh;
int AIM_bytesize, AIT_bytesize;
int expected_AIM_bytesize, expected_AIT_bytesize;
s64 AIM_byte_addr, AIT_byte_addr, fsckwsp_addr;
s64 byte_addr_diff0, byte_addr_diff1;
s32 bsize;
if ((rc = readSuper(sb, &bh)))
return rc;
j_sb = (struct jfs_superblock *)bh->b_data;
/*
* validate superblock
*/
/* validate fs signature */
if (strncmp(j_sb->s_magic, JFS_MAGIC, 4) ||
le32_to_cpu(j_sb->s_version) > JFS_VERSION) {
rc = -EINVAL;
goto out;
}
bsize = le32_to_cpu(j_sb->s_bsize);
#ifdef _JFS_4K
if (bsize != PSIZE) {
jfs_err("Currently only 4K block size supported!");
rc = -EINVAL;
goto out;
}
#endif /* _JFS_4K */
jfs_info("superblock: flag:0x%08x state:0x%08x size:0x%Lx",
le32_to_cpu(j_sb->s_flag), le32_to_cpu(j_sb->s_state),
(unsigned long long) le64_to_cpu(j_sb->s_size));
/* validate the descriptors for Secondary AIM and AIT */
if ((j_sb->s_flag & cpu_to_le32(JFS_BAD_SAIT)) !=
cpu_to_le32(JFS_BAD_SAIT)) {
expected_AIM_bytesize = 2 * PSIZE;
AIM_bytesize = lengthPXD(&(j_sb->s_aim2)) * bsize;
expected_AIT_bytesize = 4 * PSIZE;
AIT_bytesize = lengthPXD(&(j_sb->s_ait2)) * bsize;
AIM_byte_addr = addressPXD(&(j_sb->s_aim2)) * bsize;
AIT_byte_addr = addressPXD(&(j_sb->s_ait2)) * bsize;
byte_addr_diff0 = AIT_byte_addr - AIM_byte_addr;
fsckwsp_addr = addressPXD(&(j_sb->s_fsckpxd)) * bsize;
byte_addr_diff1 = fsckwsp_addr - AIT_byte_addr;
if ((AIM_bytesize != expected_AIM_bytesize) ||
(AIT_bytesize != expected_AIT_bytesize) ||
(byte_addr_diff0 != AIM_bytesize) ||
(byte_addr_diff1 <= AIT_bytesize))
j_sb->s_flag |= cpu_to_le32(JFS_BAD_SAIT);
}
if ((j_sb->s_flag & cpu_to_le32(JFS_GROUPCOMMIT)) !=
cpu_to_le32(JFS_GROUPCOMMIT))
j_sb->s_flag |= cpu_to_le32(JFS_GROUPCOMMIT);
/* validate fs state */
if (j_sb->s_state != cpu_to_le32(FM_CLEAN) &&
!(sb->s_flags & MS_RDONLY)) {
jfs_err("jfs_mount: Mount Failure: File System Dirty.");
rc = -EINVAL;
goto out;
}
sbi->state = le32_to_cpu(j_sb->s_state);
sbi->mntflag = le32_to_cpu(j_sb->s_flag);
/*
* JFS always does I/O by 4K pages. Don't tell the buffer cache
* that we use anything else (leave s_blocksize alone).
*/
sbi->bsize = bsize;
sbi->l2bsize = le16_to_cpu(j_sb->s_l2bsize);
/*
* For now, ignore s_pbsize, l2bfactor. All I/O going through buffer
* cache.
*/
sbi->nbperpage = PSIZE >> sbi->l2bsize;
sbi->l2nbperpage = L2PSIZE - sbi->l2bsize;
sbi->l2niperblk = sbi->l2bsize - L2DISIZE;
if (sbi->mntflag & JFS_INLINELOG)
sbi->logpxd = j_sb->s_logpxd;
else {
sbi->logdev = new_decode_dev(le32_to_cpu(j_sb->s_logdev));
memcpy(sbi->uuid, j_sb->s_uuid, sizeof(sbi->uuid));
memcpy(sbi->loguuid, j_sb->s_loguuid, sizeof(sbi->uuid));
}
sbi->fsckpxd = j_sb->s_fsckpxd;
sbi->ait2 = j_sb->s_ait2;
out:
brelse(bh);
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.
*
*/
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;
}
/*
* 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;
}
int jfs_mount(struct super_block *sb)
{
int rc = 0;
struct jfs_sb_info *sbi = JFS_SBI(sb);
struct inode *ipaimap = NULL;
struct inode *ipaimap2 = NULL;
struct inode *ipimap = NULL;
struct inode *ipbmap = NULL;
if ((rc = chkSuper(sb))) {
goto errout20;
}
ipaimap = diReadSpecial(sb, AGGREGATE_I, 0);
if (ipaimap == NULL) {
jfs_err("jfs_mount: Failed to read AGGREGATE_I");
rc = -EIO;
goto errout20;
}
sbi->ipaimap = ipaimap;
jfs_info("jfs_mount: ipaimap:0x%p", ipaimap);
if ((rc = diMount(ipaimap))) {
jfs_err("jfs_mount: diMount(ipaimap) failed w/rc = %d", rc);
goto errout21;
}
ipbmap = diReadSpecial(sb, BMAP_I, 0);
if (ipbmap == NULL) {
rc = -EIO;
goto errout22;
}
jfs_info("jfs_mount: ipbmap:0x%p", ipbmap);
sbi->ipbmap = ipbmap;
if ((rc = dbMount(ipbmap))) {
jfs_err("jfs_mount: dbMount failed w/rc = %d", rc);
goto errout22;
}
if ((sbi->mntflag & JFS_BAD_SAIT) == 0) {
ipaimap2 = diReadSpecial(sb, AGGREGATE_I, 1);
if (!ipaimap2) {
jfs_err("jfs_mount: Failed to read AGGREGATE_I");
rc = -EIO;
goto errout35;
}
sbi->ipaimap2 = ipaimap2;
jfs_info("jfs_mount: ipaimap2:0x%p", ipaimap2);
if ((rc = diMount(ipaimap2))) {
jfs_err("jfs_mount: diMount(ipaimap2) failed, rc = %d",
rc);
goto errout35;
}
} else
sbi->ipaimap2 = NULL;
ipimap = diReadSpecial(sb, FILESYSTEM_I, 0);
if (ipimap == NULL) {
jfs_err("jfs_mount: Failed to read FILESYSTEM_I");
rc = -EIO;
goto errout40;
}
jfs_info("jfs_mount: ipimap:0x%p", ipimap);
sbi->ipimap = ipimap;
if ((rc = diMount(ipimap))) {
jfs_err("jfs_mount: diMount failed w/rc = %d", rc);
goto errout41;
}
goto out;
errout41:
diFreeSpecial(ipimap);
errout40:
if (ipaimap2) {
diUnmount(ipaimap2, 1);
diFreeSpecial(ipaimap2);
}
errout35:
dbUnmount(ipbmap, 1);
diFreeSpecial(ipbmap);
errout22:
diUnmount(ipaimap, 1);
errout21:
diFreeSpecial(ipaimap);
errout20:
out:
if (rc)
jfs_err("Mount JFS Failure: %d", 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_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
*
*/
static 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;
jfs_info("jfs_create: dip:0x%p name:%s", 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))) {
jfs_err("jfs_create: dtSearch returned %d", 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))) {
jfs_err("jfs_create: dtInsert returned %d", 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:
jfs_info("jfs_create: rc:%d", rc);
return rc;
}
