static int tux3_symlink(struct inode *dir, struct dentry *dentry,
const char *symname)
{
struct inode *inode;
int err;
change_begin(tux_sb(dir->i_sb));
inode = tux_create_inode(dir, S_IFLNK | S_IRWXUGO, 0);
err = PTR_ERR(inode);
if (!IS_ERR(inode)) {
err = page_symlink(inode, symname, strlen(symname) + 1);
if (!err) {
err = tux_add_dirent(dir, dentry, inode);
if (!err)
goto out;
}
inode_dec_link_count(inode);
iput(inode);
}
out:
change_end(tux_sb(dir->i_sb));
return err;
}
static void tux3_write_super(struct super_block *sb)
{
struct buffer_head *bh;
BUG_ON(SB_LOC < sb->s_blocksize);
bh = sb_bread(sb, SB_LOC >> sb->s_blocksize_bits);
if (!bh) {
printk(KERN_ERR "TUX3: unable to read superblock\n");
return;
}
pack_sb(tux_sb(sb), bufdata(bh));
brelse_dirty(bh);
sb->s_dirt = 0;
}
/* Modify atom refcount */
static int atomref(struct inode *atable, atom_t atom, int use)
{
struct sb *sb = tux_sb(atable->i_sb);
unsigned shift = sb->blockbits - ATOMREF_BLKBITS;
unsigned block = sb->atomref_base + ATOMREF_SIZE * (atom >> shift);
unsigned offset = atom & ~(-1 << shift), kill = 0;
struct buffer_head *buffer;
__be16 *refcount;
int err;
buffer = blockread(mapping(atable), block);
if (!buffer)
return -EIO;
refcount = bufdata(buffer);
int low = be16_to_cpu(refcount[offset]) + use;
trace("inc atom %x by %d, offset %x[%x], low = %d",
atom, use, block, offset, low);
/* This releases buffer */
err = update_refcount(sb, buffer, offset, low);
if (err)
return err;
if (!low || (low & (-1 << 16))) {
buffer = blockread(mapping(atable), block + 1);
if (!buffer)
return -EIO;
refcount = bufdata(buffer);
int high = be16_to_cpu(refcount[offset]);
if (!low)
blockput(buffer);
else {
trace("carry %d, offset %x[%x], high = %d",
(low >> 16), block, offset, high);
high += (low >> 16);
assert(high >= 0); /* paranoia check */
/* This releases buffer */
err = update_refcount(sb, buffer, offset, high);
if (err) {
/* FIXME: better set a flag that atomref broke
* or something! */
return err;
}
}
kill = !(low | high);
}
static int __tux3_mknod(struct inode *dir, struct dentry *dentry,
struct tux_iattr *iattr, dev_t rdev)
{
if(DEBUG_MODE_K==1)
{
printf("\t\t\t\t%25s[K] %25s %4d #in\n",__FILE__,__func__,__LINE__);
}
struct inode *inode;
int err, is_dir = S_ISDIR(iattr->mode);
if (!huge_valid_dev(rdev))
return -EINVAL;
if (is_dir && dir->i_nlink >= TUX_LINK_MAX)
return -EMLINK;
change_begin(tux_sb(dir->i_sb));
inode = tux_create_inode(dir, iattr, rdev);
err = PTR_ERR(inode);
if (!IS_ERR(inode)) {
err = tux_add_dirent(dir, dentry, inode);
if (!err) {
unlock_new_inode(inode);
if (is_dir)
inode_inc_link_count(dir);
goto out;
}
clear_nlink(inode);
tux3_mark_inode_dirty(inode);
unlock_new_inode(inode);
iput(inode);
}
out:
change_end(tux_sb(dir->i_sb));
return err;
}
static void tux3_put_super(struct super_block *sb)
{
struct sb *sbi = tux_sb(sb);
/* FIXME: remove this, then use sb->s_dirt instead */
tux3_write_super(sb);
empty_stash(&sbi->defree);
iput(sbi->atable);
iput(sbi->bitmap);
iput(sbi->volmap);
iput(sbi->logmap);
sb->s_fs_info = NULL;
kfree(sbi);
}
/* Convert atom to name */
static int unatom(struct inode *atable, atom_t atom, char *name, unsigned size)
{
if(DEBUG_MODE_K==1)
{
printf("\t\t\t\t%25s[K] %25s %4d #in\n",__FILE__,__func__,__LINE__);
}
struct sb *sb = tux_sb(atable->i_sb);
struct buffer_head *buffer;
int err;
loff_t where = unatom_dict_read(atable, atom);
if (where < 0) {
err = where;
goto error;
}
buffer = blockread(mapping(atable), where >> sb->blockbits);
if (!buffer) {
err = -EIO;
goto error;
}
tux_dirent *entry = bufdata(buffer) + (where & sb->blockmask);
if (entry_atom(entry) != atom) {
tux3_fs_error(sb, "atom %x reverse entry broken", atom);
err = -EIO;
goto error_blockput;
}
unsigned len = entry->name_len;
if (size) {
if (len > size) {
err = -ERANGE;
goto error_blockput;
}
memcpy(name, entry->name, len);
}
blockput(buffer);
return len;
error_blockput:
blockput(buffer);
error:
return err;
}
/* Find atom of name */
static int find_atom(struct inode *atable, const char *name, unsigned len,
atom_t *atom)
{
struct sb *sb = tux_sb(atable->i_sb);
struct buffer_head *buffer;
struct tux3_dirent *entry;
entry = tux_find_entry(atable, name, len, &buffer, sb->atomdictsize);
if (IS_ERR(entry)) {
int err = PTR_ERR(entry);
if (err == -ENOENT)
return -ENODATA;
return err;
}
*atom = entry_atom(entry);
blockput(buffer);
return 0;
}
/*
* Mark inode dirty to delete. (called from ->drop_inode()).
* Caller must hold inode->i_lock.
*/
void tux3_mark_inode_to_delete(struct inode *inode)
{
struct sb *sb = tux_sb(inode->i_sb);
struct tux3_inode *tuxnode = tux_inode(inode);
unsigned delta;
/* inode has dead mark already */
if (tux3_inode_is_dead(tuxnode))
return;
change_begin_atomic(sb);
delta = tux3_inode_delta(inode);
__tux3_mark_inode_to_delete(inode, delta);
/*
* Hack: this is called under inode->i_lock. So, we have to
* release inode->i_lock to call mark_inode_dirty_sync().
*
* FIXME: we want to set I_DIRTY_SYNC (I_DIRTY_SYNC will
* prevent the indo is freed) and wakeup flusher if need,
* while preventing inode is freed. Need better way to do.
*/
if (!(tux3_dirty_flags(inode, delta) & I_DIRTY_SYNC)) {
/* FIXME: I_REFERENCED can't prevent completely */
//inode->i_state |= I_REFERENCED;
/* FIXME: I_WILL_FREE will bother igrab() grabs reference */
inode->i_state |= I_WILL_FREE;
spin_unlock(&inode->i_lock);
/* Tell dead inode to backend by marking as dirty. */
tux3_mark_inode_dirty_sync(inode);
spin_lock(&inode->i_lock);
inode->i_state &= ~I_WILL_FREE;
#ifdef __KERNEL__
wake_up_bit(&inode->i_state, __I_NEW);
#endif
}
change_end_atomic(sb);
}
static void tux3_put_super(struct super_block *sb)
{
struct sb *sbi = tux_sb(sb);
/* FIXME: remove this, then use sb->s_dirt instead */
tux3_write_super(sb);
destroy_defer_bfree(&sbi->new_decycle);
destroy_defer_bfree(&sbi->decycle);
destroy_defer_bfree(&sbi->derollup);
destroy_defer_bfree(&sbi->defree);
iput(sbi->atable);
iput(sbi->bitmap);
iput(sbi->volmap);
iput(sbi->logmap);
BUG_ON(!list_empty(&sbi->alloc_inodes));
sb->s_fs_info = NULL;
kfree(sbi);
}
static int tux3_statfs(struct dentry *dentry, struct kstatfs *buf)
{
struct super_block *sb = dentry->d_sb;
struct sb *sbi = tux_sb(sb);
buf->f_type = sb->s_magic;
buf->f_bsize = sbi->blocksize;
buf->f_blocks = sbi->volblocks;
buf->f_bfree = sbi->freeblocks;
buf->f_bavail = sbi->freeblocks;
#if 0
buf->f_files = buf->f_blocks << (sbi->clus_bits - EXFAT_CHUNK_BITS) / 3;
buf->f_ffree = buf->f_blocks << (sbi->clus_bits - EXFAT_CHUNK_BITS) / 3;
buf->f_fsid.val[0] = sbi->serial_number;
/*buf->f_fsid.val[1];*/
#endif
buf->f_namelen = TUX_NAME_LEN;
// buf->f_frsize = sbi->blocksize;
return 0;
}
static int __tux3_symlink(struct inode *dir, struct dentry *dentry,
struct tux_iattr *iattr, const char *symname)
{
struct sb *sb = tux_sb(dir->i_sb);
struct inode *inode;
unsigned len = strlen(symname) + 1;
int err, err2;
/* FIXME: We want more length? */
if (len > PAGE_CACHE_SIZE)
return -ENAMETOOLONG;
change_begin(sb);
inode = tux_new_inode(dir, iattr, 0);
err = PTR_ERR(inode);
if (!IS_ERR(inode)) {
err = tux_create_dirent(dir, &dentry->d_name, inode);
if (!err) {
/* FIXME: we may want to initialize symlink earlier */
err = page_symlink(inode, symname, len);
if (!err) {
d_instantiate(dentry, inode);
unlock_new_inode(inode);
goto out;
}
err2 = tux_del_dirent(dir, dentry);
if (err2)
tux3_fs_error(sb, "Failed to recover dir entry (err %d)", err2);
clear_nlink(inode);
tux3_mark_inode_dirty(inode);
unlock_new_inode(inode);
iput(inode);
}
}
out:
change_end(sb);
return err;
}
int tux_dir_is_empty(struct inode *dir)
{
struct sb *sb = tux_sb(dir->i_sb);
block_t block, blocks = dir->i_size >> sb->blockbits;
__be64 self = cpu_to_be64(tux_inode(dir)->inum);
struct buffer_head *buffer;
for (block = 0; block < blocks; block++) {
buffer = blockread(mapping(dir), block);
if (!buffer)
return -EIO;
tux_dirent *entry = bufdata(buffer);
tux_dirent *limit = bufdata(buffer) + sb->blocksize - TUX_REC_LEN(1);
for (; entry <= limit; entry = next_entry(entry)) {
if (!entry->rec_len) {
blockput(buffer);
tux_zero_len_error(dir, block);
return -EIO;
}
if (is_deleted(entry))
continue;
if (entry->name[0] != '.')
goto not_empty;
if (entry->name_len > 2)
goto not_empty;
if (entry->name_len < 2) {
if (entry->inum != self)
goto not_empty;
} else if (entry->name[1] != '.')
goto not_empty;
}
blockput(buffer);
}
return 0;
not_empty:
blockput(buffer);
return -ENOTEMPTY;
}
static struct dentry *tux3_lookup(struct inode *dir, struct dentry *dentry,
unsigned int flags)
{
struct buffer_head *buffer;
struct inode *inode;
tux_dirent *entry;
inum_t inum;
entry = tux_find_dirent(dir, &dentry->d_name, &buffer);
if (IS_ERR(entry)) {
if (PTR_ERR(entry) != -ENOENT)
return ERR_CAST(entry);
inode = NULL;
goto out;
}
inum = be64_to_cpu(entry->inum);
blockput(buffer);
inode = tux3_iget(tux_sb(dir->i_sb), inum);
out:
return d_splice_alias(inode, dentry);
}
/* Find atom of name */
static int find_atom(struct inode *atable, const char *name, unsigned len,
atom_t *atom)
{
if(DEBUG_MODE_K==1)
{
printf("\t\t\t\t%25s[K] %25s %4d #in\n",__FILE__,__func__,__LINE__);
}
struct sb *sb = tux_sb(atable->i_sb);
struct buffer_head *buffer;
tux_dirent *entry;
entry = tux_find_entry(atable, name, len, &buffer, sb->atomdictsize);
if (IS_ERR(entry)) {
int err = PTR_ERR(entry);
if (err == -ENOENT)
return -ENODATA;
return err;
}
*atom = entry_atom(entry);
blockput(buffer);
return 0;
}
/* Find free atom */
static int get_freeatom(struct inode *atable, atom_t *atom)
{
struct sb *sb = tux_sb(atable->i_sb);
atom_t freeatom = sb->freeatom;
if (!freeatom) {
*atom = sb->atomgen++;
return 0;
}
loff_t next = unatom_dict_read(atable, freeatom);
if (next < 0)
return next;
if (!is_free_unatom(next)) {
tux3_fs_error(sb, "something horrible happened");
return -EIO;
}
*atom = freeatom;
sb->freeatom = next & ~UNATOM_FREE_MASK;
return 0;
}
static int __tux3_symlink(struct inode *dir, struct dentry *dentry,
struct tux_iattr *iattr, const char *symname)
{
if(DEBUG_MODE_K==1)
{
printf("\t\t\t\t%25s[K] %25s %4d #in\n",__FILE__,__func__,__LINE__);
}
struct sb *sb = tux_sb(dir->i_sb);
struct inode *inode;
unsigned len = strlen(symname) + 1;
int err;
/* FIXME: We want more length? */
if (len > PAGE_CACHE_SIZE)
return -ENAMETOOLONG;
change_begin(sb);
inode = tux_create_inode(dir, iattr, 0);
err = PTR_ERR(inode);
if (!IS_ERR(inode)) {
err = page_symlink(inode, symname, len);
if (!err) {
err = tux_add_dirent(dir, dentry, inode);
if (!err) {
unlock_new_inode(inode);
goto out;
}
}
inode_dec_link_count(inode);
unlock_new_inode(inode);
iput(inode);
}
out:
change_end(sb);
return err;
}
/* Make atom for name */
static int make_atom(struct inode *atable, const char *name, unsigned len,
atom_t *atom)
{
if(DEBUG_MODE_K==1)
{
printf("\t\t\t\t%25s[K] %25s %4d #in\n",__FILE__,__func__,__LINE__);
}
struct sb *sb = tux_sb(atable->i_sb);
int err;
err = find_atom(atable, name, len, atom);
if (!err)
return 0;
if (err != -ENODATA)
return err;
err = get_freeatom(atable, atom);
if (err)
return err;
loff_t where = tux_create_entry(atable, name, len, *atom, 0,
&sb->atomdictsize);
if (where < 0) {
/* FIXME: better set a flag that unatom broke or something!!! */
return where;
}
/* Enter into reverse map - maybe verify zero refs? */
where = unatom_dict_write(atable, *atom, where);
if (where < 0) {
/* FIXME: better set a flag that unatom broke or something!!! */
return where;
}
return 0;
}
static int tux3_link(struct dentry *old_dentry, struct inode *dir,
struct dentry *dentry)
{
struct inode *inode = old_dentry->d_inode;
struct sb *sb = tux_sb(inode->i_sb);
int err;
if (inode->i_nlink >= TUX_LINK_MAX)
return -EMLINK;
change_begin(sb);
tux3_iattrdirty(inode);
inode->i_ctime = gettime();
inode_inc_link_count(inode);
ihold(inode);
err = tux_add_dirent(dir, dentry, inode);
if (err) {
inode_dec_link_count(inode);
iput(inode);
}
change_end(sb);
return err;
}
/* Make atom for name */
static int make_atom(struct inode *atable, const char *name, unsigned len,
atom_t *atom)
{
struct sb *sb = tux_sb(atable->i_sb);
struct buffer_head *buffer;
loff_t where;
int err;
err = find_atom(atable, name, len, atom);
if (!err)
return 0;
if (err != -ENODATA)
return err;
err = get_freeatom(atable, atom);
if (err)
return err;
where = tux_alloc_entry(atable, name, len, &sb->atomdictsize, &buffer);
if (where < 0) {
/* FIXME: better set a flag that unatom broke or something!!! */
return where;
}
/* This releases buffer */
tux_set_entry(buffer, bufdata(buffer) + (where & sb->blockmask),
*atom, 0);
/* Enter into reverse map - maybe verify zero refs? */
where = unatom_dict_write(atable, *atom, where);
if (where < 0) {
/* FIXME: better set a flag that unatom broke or something!!! */
return where;
}
return 0;
}
static int tux3_fill_super(struct super_block *sb, void *data, int silent)
{
struct sb *sbi;
int err, blocksize;
sbi = kzalloc(sizeof(struct sb), GFP_KERNEL);
if (!sbi)
return -ENOMEM;
sbi->vfs_sb = sb;
sb->s_fs_info = sbi;
sb->s_maxbytes = MAX_LFS_FILESIZE;
sb->s_magic = TUX3_SUPER_MAGIC;
sb->s_op = &tux3_super_ops;
sb->s_time_gran = 1;
mutex_init(&sbi->loglock);
INIT_LIST_HEAD(&sbi->alloc_inodes);
err = -EIO;
blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
if (!blocksize) {
if (!silent)
printk(KERN_ERR "TUX3: unable to set blocksize\n");
goto error;
}
if ((err = load_sb(tux_sb(sb)))) {
if (!silent) {
if (err == -EINVAL)
warn("invalid superblock [%Lx]",
(L)from_be_u64(*(be_u64 *)sbi->super.magic));
else
warn("Unable to read superblock");
}
goto error;
}
if (sbi->blocksize != blocksize) {
if (!sb_set_blocksize(sb, sbi->blocksize)) {
printk(KERN_ERR "TUX3: blocksize too small for device.\n");
goto error;
}
}
warn("s_blocksize %lu", sb->s_blocksize);
err = -ENOMEM;
sbi->volmap = tux_new_volmap(tux_sb(sb));
if (!sbi->volmap)
goto error;
insert_inode_hash(sbi->volmap);
sbi->logmap = tux_new_logmap(tux_sb(sb));
if (!sbi->logmap)
goto error_logmap;
err = load_itable(sbi);
if (err)
goto error_bitmap;
// struct inode *vtable;
sbi->bitmap = tux3_iget(sb, TUX_BITMAP_INO);
err = PTR_ERR(sbi->bitmap);
if (IS_ERR(sbi->bitmap))
goto error_bitmap;
sbi->rootdir = tux3_iget(sb, TUX_ROOTDIR_INO);
err = PTR_ERR(sbi->rootdir);
if (IS_ERR(sbi->rootdir))
goto error_rootdir;
sbi->atable = tux3_iget(sb, TUX_ATABLE_INO);
err = PTR_ERR(sbi->atable);
if (IS_ERR(sbi->atable))
goto error_atable;
sb->s_root = d_alloc_root(sbi->rootdir);
if (!sb->s_root)
goto error_alloc_root;
return 0;
error_alloc_root:
iput(sbi->atable);
error_atable:
iput(sbi->rootdir);
error_rootdir:
iput(sbi->bitmap);
error_bitmap:
iput(sbi->logmap);
error_logmap:
iput(sbi->volmap);
error:
kfree(sbi);
return err;
}
static int tux3_rename(struct inode *old_dir, struct dentry *old_dentry,
struct inode *new_dir, struct dentry *new_dentry)
{
struct inode *old_inode = old_dentry->d_inode;
struct inode *new_inode = new_dentry->d_inode;
struct buffer_head *old_buffer, *new_buffer;
tux_dirent *old_entry, *new_entry;
int err, new_subdir = 0;
old_entry = tux_find_dirent(old_dir, old_dentry->d_name.name,
old_dentry->d_name.len, &old_buffer);
if (IS_ERR(old_entry))
return PTR_ERR(old_entry);
/* FIXME: is this needed? */
BUG_ON(from_be_u64(old_entry->inum) != tux_inode(old_inode)->inum);
change_begin(tux_sb(old_inode->i_sb));
if (new_inode) {
int old_is_dir = S_ISDIR(old_inode->i_mode);
if (old_is_dir) {
err = tux_dir_is_empty(new_inode);
if (err)
goto error;
}
new_entry = tux_find_dirent(new_dir, new_dentry->d_name.name,
new_dentry->d_name.len, &new_buffer);
if (IS_ERR(new_entry)) {
BUG_ON(PTR_ERR(new_entry) == -ENOENT);
err = PTR_ERR(new_entry);
goto error;
}
/* this releases new_buffer */
tux_update_dirent(new_buffer, new_entry, old_inode);
new_inode->i_ctime = new_dir->i_ctime;
if (old_is_dir)
drop_nlink(new_inode);
inode_dec_link_count(new_inode);
} else {
new_subdir = S_ISDIR(old_inode->i_mode) && new_dir != old_dir;
if (new_subdir) {
if (new_dir->i_nlink >= TUX_LINK_MAX) {
err = -EMLINK;
goto error;
}
}
err = __tux_add_dirent(new_dir, new_dentry, old_inode);
if (err)
goto error;
if (new_subdir)
inode_inc_link_count(new_dir);
}
old_inode->i_ctime = new_dir->i_ctime;
mark_inode_dirty(old_inode);
err = tux_delete_dirent(old_buffer, old_entry);
if (err) {
printk(KERN_ERR "TUX3: %s: couldn't delete old entry (%Lu)\n",
__func__, (L)tux_inode(old_inode)->inum);
/* FIXME: now, we have hardlink even if it's dir. */
inode_inc_link_count(old_inode);
}
if (!err && new_subdir)
inode_dec_link_count(old_dir);
change_end(tux_sb(old_inode->i_sb));
return err;
error:
change_end(tux_sb(old_inode->i_sb));
brelse(old_buffer);
return err;
}
static int tux3_fill_super(struct super_block *sb, void *data, int silent)
{
static struct tux_iattr iattr;
struct sb *sbi;
struct root iroot;
int err, blocksize;
sbi = kzalloc(sizeof(struct sb), GFP_KERNEL);
if (!sbi)
return -ENOMEM;
sbi->vfs_sb = sb;
sb->s_fs_info = sbi;
sb->s_maxbytes = MAX_LFS_FILESIZE;
sb->s_magic = 0x54555833;
sb->s_op = &tux3_super_ops;
sb->s_time_gran = 1;
mutex_init(&sbi->loglock);
err = -EIO;
blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
if (!blocksize) {
if (!silent)
printk(KERN_ERR "TUX3: unable to set blocksize\n");
goto error;
}
err = tux_load_sb(sb, &iroot, silent);
if (err)
goto error;
printk("%s: depth %Lu, block %Lu\n",
__func__, (L)iroot.depth, (L)iroot.block);
printk("%s: blocksize %u, blockbits %u, blockmask %08x\n",
__func__, sbi->blocksize, sbi->blockbits, sbi->blockmask);
printk("%s: volblocks %Lu, freeblocks %Lu, nextalloc %Lu\n",
__func__, sbi->volblocks, sbi->freeblocks, sbi->nextalloc);
printk("%s: freeatom %u, atomgen %u\n",
__func__, sbi->freeatom, sbi->atomgen);
if (sbi->blocksize != blocksize) {
if (!sb_set_blocksize(sb, sbi->blocksize)) {
printk(KERN_ERR "TUX3: blocksize too small for device.\n");
goto error;
}
}
printk("%s: s_blocksize %lu\n", __func__, sb->s_blocksize);
err = -ENOMEM;
sbi->volmap = tux_new_volmap(tux_sb(sb));
if (!sbi->volmap)
goto error;
/* Initialize itable btree */
init_btree(itable_btree(sbi), sbi, iroot, &itable_ops);
// struct inode *vtable;
sbi->bitmap = tux3_iget(sb, TUX_BITMAP_INO);
err = PTR_ERR(sbi->bitmap);
if (IS_ERR(sbi->bitmap))
goto error_bitmap;
sbi->rootdir = tux3_iget(sb, TUX_ROOTDIR_INO);
err = PTR_ERR(sbi->rootdir);
if (IS_ERR(sbi->rootdir))
goto error_rootdir;
sbi->atable = tux3_iget(sb, TUX_ATABLE_INO);
err = PTR_ERR(sbi->atable);
if (IS_ERR(sbi->atable))
goto error_atable;
err = -ENOMEM;
sbi->logmap = tux_new_inode(sbi->rootdir, &iattr, 0);
if (!sbi->logmap)
goto error_logmap;
sb->s_root = d_alloc_root(sbi->rootdir);
if (!sb->s_root)
goto error_alloc_root;
return 0;
error_alloc_root:
iput(sbi->logmap);
error_logmap:
iput(sbi->atable);
error_atable:
iput(sbi->rootdir);
error_rootdir:
iput(sbi->bitmap);
error_bitmap:
iput(sbi->volmap);
error:
kfree(sbi);
return err;
}
static int tux3_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
{
struct inode *inode = file_inode(vma->vm_file);
struct sb *sb = tux_sb(inode->i_sb);
struct page *clone, *page = vmf->page;
void *ptr;
int ret;
sb_start_pagefault(inode->i_sb);
retry:
down_read(&tux_inode(inode)->truncate_lock);
lock_page(page);
if (page->mapping != mapping(inode)) {
unlock_page(page);
ret = VM_FAULT_NOPAGE;
goto out;
}
/*
* page fault can be happened while holding change_begin/end()
* (e.g. copy of user data between ->write_begin and
* ->write_end for write(2)).
*
* So, we use nested version here.
*/
change_begin_atomic_nested(sb, &ptr);
/*
* FIXME: Caller releases vmf->page (old_page) unconditionally.
* So, this takes additional refcount to workaround it.
*/
if (vmf->page == page)
page_cache_get(page);
clone = pagefork_for_blockdirty(page, tux3_get_current_delta());
if (IS_ERR(clone)) {
/* Someone did page fork */
pgoff_t index = page->index;
change_end_atomic_nested(sb, ptr);
unlock_page(page);
page_cache_release(page);
up_read(&tux_inode(inode)->truncate_lock);
switch (PTR_ERR(clone)) {
case -EAGAIN:
page = find_get_page(inode->i_mapping, index);
assert(page);
goto retry;
case -ENOMEM:
ret = VM_FAULT_OOM;
break;
default:
ret = VM_FAULT_SIGBUS;
break;
}
goto out;
}
file_update_time(vma->vm_file);
/* Assign buffers to dirty */
if (!page_has_buffers(clone))
create_empty_buffers(clone, sb->blocksize, 0);
/*
* We mark the page dirty already here so that when freeze is in
* progress, we are guaranteed that writeback during freezing will
* see the dirty page and writeprotect it again.
*/
tux3_set_page_dirty(clone);
#if 1
/* FIXME: Caller doesn't see the changed vmf->page */
vmf->page = clone;
change_end_atomic_nested(sb, ptr);
/* FIXME: caller doesn't know about pagefork */
unlock_page(clone);
page_cache_release(clone);
ret = 0;
// ret = VM_FAULT_LOCKED;
#endif
out:
up_read(&tux_inode(inode)->truncate_lock);
sb_end_pagefault(inode->i_sb);
return ret;
}
static int tux3_symlink(struct inode *dir, struct dentry *dentry,
const char *symname)
{
struct tux_iattr iattr = {
.uid = current_fsuid(),
.gid = current_fsgid(),
.mode = S_IFLNK | S_IRWXUGO,
};
return __tux3_symlink(dir, dentry, &iattr, symname);
}
#endif /* !__KERNEL__ */
static int tux3_unlink(struct inode *dir, struct dentry *dentry)
{
struct inode *inode = dentry->d_inode;
struct sb *sb = tux_sb(inode->i_sb);
change_begin(sb);
int err = tux_del_dirent(dir, dentry);
if (!err) {
tux3_iattrdirty(inode);
inode->i_ctime = dir->i_ctime;
/* FIXME: we shouldn't write inode for i_nlink = 0? */
inode_dec_link_count(inode);
}
change_end(sb);
return err;
}
static int tux3_rmdir(struct inode *dir, struct dentry *dentry)
{
struct sb *sb = tux_sb(dir->i_sb);
struct inode *inode = dentry->d_inode;
int err = tux_dir_is_empty(inode);
if (!err) {
change_begin(sb);
err = tux_del_dirent(dir, dentry);
if (!err) {
tux3_iattrdirty(inode);
inode->i_ctime = dir->i_ctime;
/* FIXME: we need to do this for POSIX? */
/* inode->i_size = 0; */
clear_nlink(inode);
tux3_mark_inode_dirty_sync(inode);
inode_dec_link_count(dir);
}
change_end(sb);
}
return err;
}
static int tux3_rename(struct inode *old_dir, struct dentry *old_dentry,
struct inode *new_dir, struct dentry *new_dentry)
{
struct inode *old_inode = old_dentry->d_inode;
struct inode *new_inode = new_dentry->d_inode;
struct sb *sb = tux_sb(old_inode->i_sb);
struct buffer_head *old_buffer, *new_buffer, *clone;
tux_dirent *old_entry, *new_entry;
void *olddata;
int err, new_subdir = 0;
unsigned delta;
old_entry = tux_find_dirent(old_dir, &old_dentry->d_name, &old_buffer);
if (IS_ERR(old_entry))
return PTR_ERR(old_entry);
/* FIXME: is this needed? */
assert(be64_to_cpu(old_entry->inum) == tux_inode(old_inode)->inum);
change_begin(sb);
delta = tux3_get_current_delta();
if (new_inode) {
int old_is_dir = S_ISDIR(old_inode->i_mode);
if (old_is_dir) {
err = tux_dir_is_empty(new_inode);
if (err)
goto error;
}
new_entry = tux_find_dirent(new_dir, &new_dentry->d_name,
&new_buffer);
if (IS_ERR(new_entry)) {
assert(PTR_ERR(new_entry) != -ENOENT);
err = PTR_ERR(new_entry);
goto error;
}
/*
* The directory is protected by i_mutex.
* blockdirty() should never return -EAGAIN.
*/
olddata = bufdata(new_buffer);
clone = blockdirty(new_buffer, delta);
if (IS_ERR(clone)) {
assert(PTR_ERR(clone) != -EAGAIN);
blockput(new_buffer);
err = PTR_ERR(clone);
goto error;
}
new_entry = ptr_redirect(new_entry, olddata, bufdata(clone));
/* this releases new_buffer */
tux_update_dirent(new_dir, clone, new_entry, old_inode);
tux3_iattrdirty(new_inode);
new_inode->i_ctime = new_dir->i_ctime;
if (old_is_dir)
drop_nlink(new_inode);
inode_dec_link_count(new_inode);
} else {
new_subdir = S_ISDIR(old_inode->i_mode) && new_dir != old_dir;
if (new_subdir) {
if (new_dir->i_nlink >= TUX_LINK_MAX) {
err = -EMLINK;
goto error;
}
}
err = tux_create_dirent(new_dir, &new_dentry->d_name,
old_inode);
if (err)
goto error;
if (new_subdir)
inode_inc_link_count(new_dir);
}
tux3_iattrdirty(old_inode);
old_inode->i_ctime = new_dir->i_ctime;
tux3_mark_inode_dirty(old_inode);
/*
* The new entry can be on same buffer with old_buffer, and
* may did buffer fork in the above path. So if old_buffer is
* forked buffer, we update the old_buffer in here.
*/
if (buffer_forked(old_buffer)) {
clone = blockget(mapping(old_dir), bufindex(old_buffer));
assert(clone);
old_entry = ptr_redirect(old_entry, bufdata(old_buffer),
bufdata(clone));
blockput(old_buffer);
old_buffer = clone;
}
err = tux_delete_dirent(old_dir, old_buffer, old_entry);
if (err) {
tux3_fs_error(sb, "couldn't delete old entry (%Lu)",
tux_inode(old_inode)->inum);
/* FIXME: now, we have hardlink even if it's dir. */
inode_inc_link_count(old_inode);
}
if (!err && new_subdir)
inode_dec_link_count(old_dir);
change_end(sb);
return err;
error:
change_end(sb);
blockput(old_buffer);
return err;
}
#ifdef __KERNEL__
const struct file_operations tux_dir_fops = {
.llseek = generic_file_llseek,
.read = generic_read_dir,
.readdir = tux_readdir,
.fsync = tux3_sync_file,
};
const struct inode_operations tux_dir_iops = {
.create = tux3_create,
.lookup = tux3_lookup,
.link = tux3_link,
.unlink = tux3_unlink,
.symlink = tux3_symlink,
.mkdir = tux3_mkdir,
.rmdir = tux3_rmdir,
.mknod = tux3_mknod,
.rename = tux3_rename,
.setattr = tux3_setattr,
.getattr = tux3_getattr
// .setxattr = generic_setxattr,
// .getxattr = generic_getxattr,
// .listxattr = ext3_listxattr,
// .removexattr = generic_removexattr,
// .permission = ext3_permission,
/* FIXME: why doesn't ext4 support this for directory? */
// .fallocate = ext4_fallocate,
// .fiemap = ext4_fiemap,
};
int main(int argc, char *argv[])
{
unsigned abits = DATA_BTREE_BIT|CTIME_SIZE_BIT|MODE_OWNER_BIT|LINK_COUNT_BIT|MTIME_BIT;
struct dev *dev = &(struct dev){ .bits = 8, .fd = open(argv[1], O_CREAT|O_RDWR, S_IRUSR|S_IWUSR) };
assert(!ftruncate(dev->fd, 1 << 24));
init_buffers(dev, 1 << 20, 0);
struct sb *sb = rapid_sb(dev,
.version = 0,
.atomref_base = 1 << 10,
.unatom_base = 1 << 11,
.atomgen = 1);
struct inode *inode = rapid_open_inode(sb, NULL, S_IFDIR | 0x666,
.present = abits, .i_uid = 0x12121212, .i_gid = 0x34343434,
.i_ctime = spectime(0xdec0debeadULL),
.i_mtime = spectime(0xbadfaced00dULL));
inode->btree = (struct btree){
.root = { .block = 0xcaba1f00dULL, .depth = 3 }
};
sb->atable = inode;
for (int i = 0; i < 2; i++) {
struct buffer_head *buffer = blockget(mapping(inode), tux_sb(inode->i_sb)->atomref_base + i);
memset(bufdata(buffer), 0, sb->blocksize);
blockput_dirty(buffer);
}
if (1) {
warn("---- test positive and negative refcount carry ----");
use_atom(inode, 6, 1 << 15);
use_atom(inode, 6, (1 << 15));
use_atom(inode, 6, -(1 << 15));
use_atom(inode, 6, -(1 << 15));
}
warn("---- test atom table ----");
printf("atom = %Lx\n", (L)make_atom(inode, "foo", 3));
printf("atom = %Lx\n", (L)make_atom(inode, "foo", 3));
printf("atom = %Lx\n", (L)make_atom(inode, "bar", 3));
printf("atom = %Lx\n", (L)make_atom(inode, "foo", 3));
printf("atom = %Lx\n", (L)make_atom(inode, "bar", 3));
warn("---- test inode xattr cache ----");
int err;
err = xcache_update(inode, 0x666, "hello", 5, 0);
if (err)
printf("err %d\n", err);
err = xcache_update(inode, 0x777, "world!", 6, 0);
if (err)
printf("err %d\n", err);
xcache_dump(inode);
struct xattr *xattr = xcache_lookup(tux_inode(inode)->xcache, 0x777);
if (!IS_ERR(xattr))
printf("atom %x => %.*s\n", xattr->atom, xattr->size, xattr->body);
err = xcache_update(inode, 0x111, "class", 5, 0);
if (err)
printf("err %d\n", err);
err = xcache_update(inode, 0x666, NULL, 0, 0);
if (err)
printf("err %d\n", err);
err = xcache_update(inode, 0x222, "boooyah", 7, 0);
if (err)
printf("err %d\n", err);
xcache_dump(inode);
warn("---- test xattr inode table encode and decode ----");
char attrs[1000] = { };
char *top = encode_xattrs(inode, attrs, sizeof(attrs));
hexdump(attrs, top - attrs);
printf("predicted size = %x, encoded size = %Lx\n", encode_xsize(inode), (L)(top - attrs));
inode->xcache->size = offsetof(struct xcache, xattrs);
char *newtop = decode_attrs(inode, attrs, top - attrs);
printf("predicted size = %x, xcache size = %x\n", decode_xsize(inode, attrs, top - attrs), inode->xcache->size);
assert(top == newtop);
xcache_dump(inode);
free(inode->xcache);
inode->xcache = NULL;
warn("---- xattr update ----");
set_xattr(inode, "hello", 5, "world!", 6, 0);
set_xattr(inode, "empty", 5, "zot", 0, 0);
set_xattr(inode, "foo", 3, "foobar", 6, 0);
xcache_dump(inode);
warn("---- xattr remove ----");
// del_xattr(inode, "hello", 5);
xcache_dump(inode);
warn("---- xattr lookup ----");
for (int i = 0, len; i < 3; i++) {
char *namelist[] = { "hello", "foo", "world" }, *name = namelist[i];
char data[100];
int size = get_xattr(inode, name, len = strlen(name), data, sizeof(data));
if (size < 0)
printf("xattr %.*s not found (%s)\n", len, name, strerror(-size));
else
printf("found xattr %.*s => %.*s\n", len, name, size, data);
}
warn("---- list xattrs ----");
int len = xattr_list(inode, attrs, sizeof(attrs));
printf("xattr list length = %i\n", xattr_list(inode, NULL, 0));
hexdump(attrs, len);
warn("---- atom reverse map ----");
for (int i = 0; i < 5; i++) {
unsigned atom = i, offset;
struct buffer_head *buffer = blockread_unatom(inode, atom, &offset);
loff_t where = from_be_u64(((be_u64 *)bufdata(buffer))[offset]);
blockput_dirty(buffer);
buffer = blockread(mapping(inode), where >> sb->blockbits);
printf("atom %.3Lx at dirent %.4Lx, ", (L)atom, (L)where);
hexdump(bufdata(buffer) + (where & sb->blockmask), 16);
blockput(buffer);
}
warn("---- atom recycle ----");
set_xattr(inode, "hello", 5, NULL, 0, 0);
show_freeatoms(sb);
printf("got free atom %x\n", get_freeatom(inode));
printf("got free atom %x\n", get_freeatom(inode));
printf("got free atom %x\n", get_freeatom(inode));
warn("---- dump atom table ----");
dump_atoms(inode);
show_buffers(inode->map);
exit(0);
}
loff_t tux_create_entry(struct inode *dir, const char *name, unsigned len,
inum_t inum, umode_t mode, loff_t *size)
{
unsigned delta = tux3_get_current_delta();
struct sb *sb = tux_sb(dir->i_sb);
tux_dirent *entry;
struct buffer_head *buffer, *clone;
unsigned reclen = TUX_REC_LEN(len), rec_len, name_len, offset;
unsigned blocksize = sb->blocksize;
block_t block, blocks = *size >> sb->blockbits;
void *olddata;
for (block = 0; block < blocks; block++) {
buffer = blockread(mapping(dir), block);
if (!buffer)
return -EIO;
entry = bufdata(buffer);
tux_dirent *limit = bufdata(buffer) + blocksize - reclen;
while (entry <= limit) {
if (entry->rec_len == 0) {
blockput(buffer);
tux_zero_len_error(dir, block);
return -EIO;
}
name_len = TUX_REC_LEN(entry->name_len);
rec_len = tux_rec_len_from_disk(entry->rec_len);
if (is_deleted(entry) && rec_len >= reclen)
goto create;
if (rec_len >= name_len + reclen)
goto create;
entry = (void *)entry + rec_len;
}
blockput(buffer);
}
entry = NULL;
buffer = blockget(mapping(dir), block);
assert(!buffer_dirty(buffer));
create:
/*
* The directory is protected by i_mutex.
* blockdirty() should never return -EAGAIN.
*/
olddata = bufdata(buffer);
clone = blockdirty(buffer, delta);
if (IS_ERR(clone)) {
assert(PTR_ERR(clone) != -EAGAIN);
blockput(buffer);
return PTR_ERR(clone);
}
if (!entry) {
/* Expanding the directory size. Initialize block. */
entry = bufdata(clone);
memset(entry, 0, blocksize);
entry->rec_len = tux_rec_len_to_disk(blocksize);
assert(is_deleted(entry));
*size += blocksize;
} else {
entry = ptr_redirect(entry, olddata, bufdata(clone));
if (!is_deleted(entry)) {
tux_dirent *newent = (void *)entry + name_len;
unsigned rest_rec_len = rec_len - name_len;
newent->rec_len = tux_rec_len_to_disk(rest_rec_len);
entry->rec_len = tux_rec_len_to_disk(name_len);
entry = newent;
}
}
entry->name_len = len;
memcpy(entry->name, name, len);
offset = (void *)entry - bufdata(clone);
/* this releases buffer */
tux_update_entry(clone, entry, inum, mode);
return (block << sb->blockbits) + offset; /* only for xattr create */
}
/*
* Almost copy of generic_file_splice_write() (added changed_begin/end,
* tux3_iattrdirty()).
*/
static ssize_t tux3_file_splice_write(struct pipe_inode_info *pipe,
struct file *out, loff_t *ppos,
size_t len, unsigned int flags)
{
if(DEBUG_MODE_K==1)
{
printk(KERN_INFO"%25s %25s %4d #in\n",__FILE__,__func__,__LINE__);
}
struct address_space *mapping = out->f_mapping;
struct inode *inode = mapping->host;
struct sb *sb = tux_sb(inode->i_sb);
struct splice_desc sd = {
.total_len = len,
.flags = flags,
.pos = *ppos,
.u.file = out,
};
ssize_t ret;
sb_start_write(inode->i_sb);
pipe_lock(pipe);
splice_from_pipe_begin(&sd);
do {
ret = splice_from_pipe_next(pipe, &sd);
if (ret <= 0)
break;
mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
/* For each ->write_end() calls change_end(). */
change_begin(sb);
/* For timestamp. FIXME: convert this to ->update_time
* handler? */
tux3_iattrdirty(inode);
ret = file_remove_suid(out);
if (!ret) {
ret = file_update_time(out);
if (!ret)
ret = splice_from_pipe_feed(pipe, &sd,
pipe_to_file);
}
change_end_if_needed(sb);
mutex_unlock(&inode->i_mutex);
} while (ret > 0);
splice_from_pipe_end(pipe, &sd);
pipe_unlock(pipe);
if (sd.num_spliced)
ret = sd.num_spliced;
if (ret > 0) {
int err;
err = generic_write_sync(out, *ppos, ret);
if (err)
ret = err;
else
*ppos += ret;
balance_dirty_pages_ratelimited(mapping);
}
sb_end_write(inode->i_sb);
return ret;
}