int jfs_umount(struct super_block *sb)
{
struct jfs_sb_info *sbi = JFS_SBI(sb);
struct inode *ipbmap = sbi->ipbmap;
struct inode *ipimap = sbi->ipimap;
struct inode *ipaimap = sbi->ipaimap;
struct inode *ipaimap2 = sbi->ipaimap2;
struct jfs_log *log;
int rc = 0;
jfs_info("UnMount JFS: sb:0x%p", sb);
if ((log = sbi->log))
/*
* Wait for outstanding transactions to be written to log:
*/
jfs_flush_journal(log, 2);
diUnmount(ipimap, 0);
diFreeSpecial(ipimap);
sbi->ipimap = NULL;
ipaimap2 = sbi->ipaimap2;
if (ipaimap2) {
diUnmount(ipaimap2, 0);
diFreeSpecial(ipaimap2);
sbi->ipaimap2 = NULL;
}
ipaimap = sbi->ipaimap;
diUnmount(ipaimap, 0);
diFreeSpecial(ipaimap);
sbi->ipaimap = NULL;
dbUnmount(ipbmap, 0);
diFreeSpecial(ipbmap);
sbi->ipimap = NULL;
filemap_write_and_wait(sbi->direct_inode->i_mapping);
if (log) {
updateSuper(sb, FM_CLEAN);
rc = lmLogClose(sb);
}
jfs_info("UnMount JFS Complete: rc = %d", rc);
return rc;
}
void jfs_delete_inode(struct inode *inode)
{
jfs_info("In jfs_delete_inode, inode = 0x%p", inode);
if (!is_bad_inode(inode))
dquot_initialize(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_initialize(inode);
dquot_free_inode(inode);
dquot_drop(inode);
}
clear_inode(inode);
}
void jfs_evict_inode(struct inode *inode)
{
jfs_info("In jfs_evict_inode, inode = 0x%p", inode);
if (!inode->i_nlink && !is_bad_inode(inode)) {
dquot_initialize(inode);
if (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_initialize(inode);
dquot_free_inode(inode);
}
} else {
truncate_inode_pages(&inode->i_data, 0);
}
end_writeback(inode);
dquot_drop(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));
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;
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);
}
void jfs_truncate(struct inode *ip)
{
jfs_info("jfs_truncate: size = 0x%lx", (ulong) ip->i_size);
nobh_truncate_page(ip->i_mapping, ip->i_size, jfs_get_block);
IWRITE_LOCK(ip, RDWRLOCK_NORMAL);
jfs_truncate_nolock(ip, ip->i_size);
IWRITE_UNLOCK(ip);
}
/*
* NAME: commitZeroLink()
*
* FUNCTION: for non-directory, called by jfs_remove(),
* truncate a regular file, directory or symbolic
* link to zero length. return 0 if type is not
* one of these.
*
* if the file is currently associated with a VM segment
* only permanent disk and inode map resources are freed,
* and neither the inode nor indirect blocks are modified
* so that the resources can be later freed in the work
* map by ctrunc1.
* if there is no VM segment on entry, the resources are
* freed in both work and permanent map.
* (? for temporary file - memory object is cached even
* after no reference:
* reference count > 0 - )
*
* PARAMETERS: cd - pointer to commit data structure.
* current inode is the one to truncate.
*
* RETURN: Errors from subroutines
*/
static s64 commitZeroLink(tid_t tid, struct inode *ip)
{
int filetype;
struct tblock *tblk;
jfs_info("commitZeroLink: tid = %d, ip = 0x%p", tid, ip);
filetype = ip->i_mode & S_IFMT;
switch (filetype) {
case S_IFREG:
break;
case S_IFLNK:
/* fast symbolic link */
if (ip->i_size < IDATASIZE) {
ip->i_size = 0;
return 0;
}
break;
default:
assert(filetype != S_IFDIR);
return 0;
}
set_cflag(COMMIT_Freewmap, ip);
/* mark transaction of block map update type */
tblk = tid_to_tblock(tid);
tblk->xflag |= COMMIT_PMAP;
/*
* free EA
*/
if (JFS_IP(ip)->ea.flag & DXD_EXTENT)
/* acquire maplock on EA to be freed from block map */
txEA(tid, ip, &JFS_IP(ip)->ea, NULL);
/*
* free ACL
*/
if (JFS_IP(ip)->acl.flag & DXD_EXTENT)
/* acquire maplock on EA to be freed from block map */
txEA(tid, ip, &JFS_IP(ip)->acl, NULL);
/*
* free xtree/data (truncate to zero length):
* free xtree/data pages from cache if COMMIT_PWMAP,
* free xtree/data blocks from persistent block map, and
* free xtree/data blocks from working block map if COMMIT_PWMAP;
*/
if (ip->i_size)
return xtTruncate_pmap(tid, ip, 0);
return 0;
}
void jfs_delete_inode(struct inode *inode)
{
jfs_info("In jfs_delete_inode, inode = 0x%p", inode);
if (test_cflag(COMMIT_Freewmap, inode))
freeZeroLink(inode);
diFree(inode);
clear_inode(inode);
}
static void txLockFree(lid_t lid)
{
TxLock[lid].tid = 0;
TxLock[lid].next = TxAnchor.freelock;
TxAnchor.freelock = lid;
TxAnchor.tlocksInUse--;
if (jfs_tlocks_low && (TxAnchor.tlocksInUse < TxLockLWM)) {
jfs_info("txLockFree jfs_tlocks_low no more");
jfs_tlocks_low = 0;
TXN_WAKEUP(&TxAnchor.lowlockwait);
}
TXN_WAKEUP(&TxAnchor.freelockwait);
}
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_delete_inode(struct inode *inode)
{
jfs_info("In jfs_delete_inode, inode = 0x%p", inode);
if (test_cflag(COMMIT_Freewmap, inode))
freeZeroLink(inode);
diFree(inode);
/*
* Free the inode from the quota allocation.
*/
DQUOT_INIT(inode);
DQUOT_FREE_INODE(inode);
DQUOT_DROP(inode);
clear_inode(inode);
}
void jfs_read_inode(struct inode *inode)
{
int rc;
rc = alloc_jfs_inode(inode);
if (rc) {
jfs_warn("In jfs_read_inode, alloc_jfs_inode failed");
goto bad_inode;
}
jfs_info("In jfs_read_inode, inode = 0x%p", inode);
if (diRead(inode))
goto bad_inode_free;
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;
inode->i_mapping->a_ops = &jfs_aops;
inode->i_mapping->gfp_mask = GFP_NOFS;
} 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_symlink_inode_operations;
} else {
inode->i_op = &jfs_file_inode_operations;
init_special_inode(inode, inode->i_mode,
kdev_t_to_nr(inode->i_rdev));
}
return;
bad_inode_free:
free_jfs_inode(inode);
bad_inode:
make_bad_inode(inode);
}
/*
* 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;
}
/*
* Get a transaction lock from the free list. If the number in use is
* greater than the high water mark, wake up the sync daemon. This should
* free some anonymous transaction locks. (TXN_LOCK must be held.)
*/
static lid_t txLockAlloc(void)
{
lid_t lid;
INCREMENT(TxStat.txLockAlloc);
if (!TxAnchor.freelock) {
INCREMENT(TxStat.txLockAlloc_freelock);
}
while (!(lid = TxAnchor.freelock))
TXN_SLEEP(&TxAnchor.freelockwait);
TxAnchor.freelock = TxLock[lid].next;
HIGHWATERMARK(stattx.maxlid, lid);
if ((++TxAnchor.tlocksInUse > TxLockHWM) && (jfs_tlocks_low == 0)) {
jfs_info("txLockAlloc tlocks low");
jfs_tlocks_low = 1;
wake_up_process(jfsSyncThread);
}
return lid;
}
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);
}
/*
* 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;
}
/*
* 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;
}
/*
* 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;
}
/*
* 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;
}
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;
}
/*
* 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;
}
/*
* 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_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: jfs_umount(vfsp, flags, crp)
*
* FUNCTION: vfs_umount()
*
* PARAMETERS: vfsp - virtual file system pointer
* flags - unmount for shutdown
* crp - credential
*
* RETURN : EBUSY - device has open files
*/
int jfs_umount(struct super_block *sb)
{
struct jfs_sb_info *sbi = JFS_SBI(sb);
struct inode *ipbmap = sbi->ipbmap;
struct inode *ipimap = sbi->ipimap;
struct inode *ipaimap = sbi->ipaimap;
struct inode *ipaimap2 = sbi->ipaimap2;
struct jfs_log *log;
int rc = 0;
jfs_info("UnMount JFS: sb:0x%p", sb);
/*
* update superblock and close log
*
* if mounted read-write and log based recovery was enabled
*/
if ((log = sbi->log))
/*
* Wait for outstanding transactions to be written to log:
*/
jfs_flush_journal(log, 1);
/*
* close fileset inode allocation map (aka fileset inode)
*/
diUnmount(ipimap, 0);
diFreeSpecial(ipimap);
sbi->ipimap = NULL;
/*
* close secondary aggregate inode allocation map
*/
ipaimap2 = sbi->ipaimap2;
if (ipaimap2) {
diUnmount(ipaimap2, 0);
diFreeSpecial(ipaimap2);
sbi->ipaimap2 = NULL;
}
/*
* close aggregate inode allocation map
*/
ipaimap = sbi->ipaimap;
diUnmount(ipaimap, 0);
diFreeSpecial(ipaimap);
sbi->ipaimap = NULL;
/*
* close aggregate block allocation map
*/
dbUnmount(ipbmap, 0);
diFreeSpecial(ipbmap);
sbi->ipimap = NULL;
/*
* Make sure all metadata makes it to disk before we mark
* the superblock as clean
*/
filemap_write_and_wait(sbi->direct_inode->i_mapping);
/*
* ensure all file system file pages are propagated to their
* home blocks on disk (and their in-memory buffer pages are
* invalidated) BEFORE updating file system superblock state
* (to signify file system is unmounted cleanly, and thus in
* consistent state) and log superblock active file system
* list (to signify skip logredo()).
*/
if (log) { /* log = NULL if read-only mount */
updateSuper(sb, FM_CLEAN);
/*
* close log:
*
* remove file system from log active file system list.
*/
rc = lmLogClose(sb);
}
jfs_info("UnMount JFS Complete: 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.
*
*/
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;
}
/*
* 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_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;
}