[net-next-2.6.git] / fs / omfs / inode.c

/*
 * Optimized MPEG FS - inode and super operations.
 * Copyright (C) 2006 Bob Copeland <me@bobcopeland.com>
 * Released under GPL v2.
 */
#include <linux/module.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/fs.h>
#include <linux/vfs.h>
#include <linux/parser.h>
#include <linux/buffer_head.h>
#include <linux/vmalloc.h>
#include <linux/writeback.h>
#include <linux/crc-itu-t.h>
#include "omfs.h"

MODULE_AUTHOR("Bob Copeland <me@bobcopeland.com>");
MODULE_DESCRIPTION("OMFS (ReplayTV/Karma) Filesystem for Linux");
MODULE_LICENSE("GPL");

struct inode *omfs_new_inode(struct inode *dir, int mode)
{
	struct inode *inode;
	u64 new_block;
	int err;
	int len;
	struct omfs_sb_info *sbi = OMFS_SB(dir->i_sb);

	inode = new_inode(dir->i_sb);
	if (!inode)
		return ERR_PTR(-ENOMEM);

	err = omfs_allocate_range(dir->i_sb, sbi->s_mirrors, sbi->s_mirrors,
			&new_block, &len);
	if (err)
		goto fail;

	inode->i_ino = new_block;
	inode_init_owner(inode, NULL, mode);
	inode->i_mapping->a_ops = &omfs_aops;

	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
	switch (mode & S_IFMT) {
	case S_IFDIR:
		inode->i_op = &omfs_dir_inops;
		inode->i_fop = &omfs_dir_operations;
		inode->i_size = sbi->s_sys_blocksize;
		inc_nlink(inode);
		break;
	case S_IFREG:
		inode->i_op = &omfs_file_inops;
		inode->i_fop = &omfs_file_operations;
		inode->i_size = 0;
		break;
	}

	insert_inode_hash(inode);
	mark_inode_dirty(inode);
	return inode;
fail:
	make_bad_inode(inode);
	iput(inode);
	return ERR_PTR(err);
}

/*
 * Update the header checksums for a dirty inode based on its contents.
 * Caller is expected to hold the buffer head underlying oi and mark it
 * dirty.
 */
static void omfs_update_checksums(struct omfs_inode *oi)
{
	int xor, i, ofs = 0, count;
	u16 crc = 0;
	unsigned char *ptr = (unsigned char *) oi;

	count = be32_to_cpu(oi->i_head.h_body_size);
	ofs = sizeof(struct omfs_header);

	crc = crc_itu_t(crc, ptr + ofs, count);
	oi->i_head.h_crc = cpu_to_be16(crc);

	xor = ptr[0];
	for (i = 1; i < OMFS_XOR_COUNT; i++)
		xor ^= ptr[i];

	oi->i_head.h_check_xor = xor;
}

static int __omfs_write_inode(struct inode *inode, int wait)
{
	struct omfs_inode *oi;
	struct omfs_sb_info *sbi = OMFS_SB(inode->i_sb);
	struct buffer_head *bh, *bh2;
	unsigned int block;
	u64 ctime;
	int i;
	int ret = -EIO;
	int sync_failed = 0;

	/* get current inode since we may have written sibling ptrs etc. */
	block = clus_to_blk(sbi, inode->i_ino);
	bh = sb_bread(inode->i_sb, block);
	if (!bh)
		goto out;

	oi = (struct omfs_inode *) bh->b_data;

	oi->i_head.h_self = cpu_to_be64(inode->i_ino);
	if (S_ISDIR(inode->i_mode))
		oi->i_type = OMFS_DIR;
	else if (S_ISREG(inode->i_mode))
		oi->i_type = OMFS_FILE;
	else {
		printk(KERN_WARNING "omfs: unknown file type: %d\n",
			inode->i_mode);
		goto out_brelse;
	}

	oi->i_head.h_body_size = cpu_to_be32(sbi->s_sys_blocksize -
		sizeof(struct omfs_header));
	oi->i_head.h_version = 1;
	oi->i_head.h_type = OMFS_INODE_NORMAL;
	oi->i_head.h_magic = OMFS_IMAGIC;
	oi->i_size = cpu_to_be64(inode->i_size);

	ctime = inode->i_ctime.tv_sec * 1000LL +
		((inode->i_ctime.tv_nsec + 999)/1000);
	oi->i_ctime = cpu_to_be64(ctime);

	omfs_update_checksums(oi);

	mark_buffer_dirty(bh);
	if (wait) {
		sync_dirty_buffer(bh);
		if (buffer_req(bh) && !buffer_uptodate(bh))
			sync_failed = 1;
	}

	/* if mirroring writes, copy to next fsblock */
	for (i = 1; i < sbi->s_mirrors; i++) {
		bh2 = sb_bread(inode->i_sb, block + i *
			(sbi->s_blocksize / sbi->s_sys_blocksize));
		if (!bh2)
			goto out_brelse;

		memcpy(bh2->b_data, bh->b_data, bh->b_size);
		mark_buffer_dirty(bh2);
		if (wait) {
			sync_dirty_buffer(bh2);
			if (buffer_req(bh2) && !buffer_uptodate(bh2))
				sync_failed = 1;
		}
		brelse(bh2);
	}
	ret = (sync_failed) ? -EIO : 0;
out_brelse:
	brelse(bh);
out:
	return ret;
}

static int omfs_write_inode(struct inode *inode, struct writeback_control *wbc)
{
	return __omfs_write_inode(inode, wbc->sync_mode == WB_SYNC_ALL);
}

int omfs_sync_inode(struct inode *inode)
{
	return __omfs_write_inode(inode, 1);
}

/*
 * called when an entry is deleted, need to clear the bits in the
 * bitmaps.
 */
static void omfs_delete_inode(struct inode *inode)
{
	truncate_inode_pages(&inode->i_data, 0);

	if (S_ISREG(inode->i_mode)) {
		inode->i_size = 0;
		omfs_shrink_inode(inode);
	}

	omfs_clear_range(inode->i_sb, inode->i_ino, 2);
	clear_inode(inode);
}

struct inode *omfs_iget(struct super_block *sb, ino_t ino)
{
	struct omfs_sb_info *sbi = OMFS_SB(sb);
	struct omfs_inode *oi;
	struct buffer_head *bh;
	unsigned int block;
	u64 ctime;
	unsigned long nsecs;
	struct inode *inode;

	inode = iget_locked(sb, ino);
	if (!inode)
		return ERR_PTR(-ENOMEM);
	if (!(inode->i_state & I_NEW))
		return inode;

	block = clus_to_blk(sbi, ino);
	bh = sb_bread(inode->i_sb, block);
	if (!bh)
		goto iget_failed;

	oi = (struct omfs_inode *)bh->b_data;

	/* check self */
	if (ino != be64_to_cpu(oi->i_head.h_self))
		goto fail_bh;

	inode->i_uid = sbi->s_uid;
	inode->i_gid = sbi->s_gid;

	ctime = be64_to_cpu(oi->i_ctime);
	nsecs = do_div(ctime, 1000) * 1000L;

	inode->i_atime.tv_sec = ctime;
	inode->i_mtime.tv_sec = ctime;
	inode->i_ctime.tv_sec = ctime;
	inode->i_atime.tv_nsec = nsecs;
	inode->i_mtime.tv_nsec = nsecs;
	inode->i_ctime.tv_nsec = nsecs;

	inode->i_mapping->a_ops = &omfs_aops;

	switch (oi->i_type) {
	case OMFS_DIR:
		inode->i_mode = S_IFDIR | (S_IRWXUGO & ~sbi->s_dmask);
		inode->i_op = &omfs_dir_inops;
		inode->i_fop = &omfs_dir_operations;
		inode->i_size = sbi->s_sys_blocksize;
		inc_nlink(inode);
		break;
	case OMFS_FILE:
		inode->i_mode = S_IFREG | (S_IRWXUGO & ~sbi->s_fmask);
		inode->i_fop = &omfs_file_operations;
		inode->i_size = be64_to_cpu(oi->i_size);
		break;
	}
	brelse(bh);
	unlock_new_inode(inode);
	return inode;
fail_bh:
	brelse(bh);
iget_failed:
	iget_failed(inode);
	return ERR_PTR(-EIO);
}

static void omfs_put_super(struct super_block *sb)
{
	struct omfs_sb_info *sbi = OMFS_SB(sb);
	kfree(sbi->s_imap);
	kfree(sbi);
	sb->s_fs_info = NULL;
}

static int omfs_statfs(struct dentry *dentry, struct kstatfs *buf)
{
	struct super_block *s = dentry->d_sb;
	struct omfs_sb_info *sbi = OMFS_SB(s);
	u64 id = huge_encode_dev(s->s_bdev->bd_dev);

	buf->f_type = OMFS_MAGIC;
	buf->f_bsize = sbi->s_blocksize;
	buf->f_blocks = sbi->s_num_blocks;
	buf->f_files = sbi->s_num_blocks;
	buf->f_namelen = OMFS_NAMELEN;
	buf->f_fsid.val[0] = (u32)id;
	buf->f_fsid.val[1] = (u32)(id >> 32);

	buf->f_bfree = buf->f_bavail = buf->f_ffree =
		omfs_count_free(s);

	return 0;
}

static const struct super_operations omfs_sops = {
	.write_inode	= omfs_write_inode,
	.delete_inode	= omfs_delete_inode,
	.put_super	= omfs_put_super,
	.statfs		= omfs_statfs,
	.show_options	= generic_show_options,
};

/*
 * For Rio Karma, there is an on-disk free bitmap whose location is
 * stored in the root block.  For ReplayTV, there is no such free bitmap
 * so we have to walk the tree.  Both inodes and file data are allocated
 * from the same map.  This array can be big (300k) so we allocate
 * in units of the blocksize.
 */
static int omfs_get_imap(struct super_block *sb)
{
	int bitmap_size;
	int array_size;
	int count;
	struct omfs_sb_info *sbi = OMFS_SB(sb);
	struct buffer_head *bh;
	unsigned long **ptr;
	sector_t block;

	bitmap_size = DIV_ROUND_UP(sbi->s_num_blocks, 8);
	array_size = DIV_ROUND_UP(bitmap_size, sb->s_blocksize);

	if (sbi->s_bitmap_ino == ~0ULL)
		goto out;

	sbi->s_imap_size = array_size;
	sbi->s_imap = kzalloc(array_size * sizeof(unsigned long *), GFP_KERNEL);
	if (!sbi->s_imap)
		goto nomem;

	block = clus_to_blk(sbi, sbi->s_bitmap_ino);
	ptr = sbi->s_imap;
	for (count = bitmap_size; count > 0; count -= sb->s_blocksize) {
		bh = sb_bread(sb, block++);
		if (!bh)
			goto nomem_free;
		*ptr = kmalloc(sb->s_blocksize, GFP_KERNEL);
		if (!*ptr) {
			brelse(bh);
			goto nomem_free;
		}
		memcpy(*ptr, bh->b_data, sb->s_blocksize);
		if (count < sb->s_blocksize)
			memset((void *)*ptr + count, 0xff,
				sb->s_blocksize - count);
		brelse(bh);
		ptr++;
	}
out:
	return 0;

nomem_free:
	for (count = 0; count < array_size; count++)
		kfree(sbi->s_imap[count]);

	kfree(sbi->s_imap);
nomem:
	sbi->s_imap = NULL;
	sbi->s_imap_size = 0;
	return -ENOMEM;
}

enum {
	Opt_uid, Opt_gid, Opt_umask, Opt_dmask, Opt_fmask
};

static const match_table_t tokens = {
	{Opt_uid, "uid=%u"},
	{Opt_gid, "gid=%u"},
	{Opt_umask, "umask=%o"},
	{Opt_dmask, "dmask=%o"},
	{Opt_fmask, "fmask=%o"},
};

static int parse_options(char *options, struct omfs_sb_info *sbi)
{
	char *p;
	substring_t args[MAX_OPT_ARGS];
	int option;

	if (!options)
		return 1;

	while ((p = strsep(&options, ",")) != NULL) {
		int token;
		if (!*p)
			continue;

		token = match_token(p, tokens, args);
		switch (token) {
		case Opt_uid:
			if (match_int(&args[0], &option))
				return 0;
			sbi->s_uid = option;
			break;
		case Opt_gid:
			if (match_int(&args[0], &option))
				return 0;
			sbi->s_gid = option;
			break;
		case Opt_umask:
			if (match_octal(&args[0], &option))
				return 0;
			sbi->s_fmask = sbi->s_dmask = option;
			break;
		case Opt_dmask:
			if (match_octal(&args[0], &option))
				return 0;
			sbi->s_dmask = option;
			break;
		case Opt_fmask:
			if (match_octal(&args[0], &option))
				return 0;
			sbi->s_fmask = option;
			break;
		default:
			return 0;
		}
	}
	return 1;
}

static int omfs_fill_super(struct super_block *sb, void *data, int silent)
{
	struct buffer_head *bh, *bh2;
	struct omfs_super_block *omfs_sb;
	struct omfs_root_block *omfs_rb;
	struct omfs_sb_info *sbi;
	struct inode *root;
	sector_t start;
	int ret = -EINVAL;

	save_mount_options(sb, (char *) data);

	sbi = kzalloc(sizeof(struct omfs_sb_info), GFP_KERNEL);
	if (!sbi)
		return -ENOMEM;

	sb->s_fs_info = sbi;

	sbi->s_uid = current_uid();
	sbi->s_gid = current_gid();
	sbi->s_dmask = sbi->s_fmask = current_umask();

	if (!parse_options((char *) data, sbi))
		goto end;

	sb->s_maxbytes = 0xffffffff;

	sb_set_blocksize(sb, 0x200);

	bh = sb_bread(sb, 0);
	if (!bh)
		goto end;

	omfs_sb = (struct omfs_super_block *)bh->b_data;

	if (omfs_sb->s_magic != cpu_to_be32(OMFS_MAGIC)) {
		if (!silent)
			printk(KERN_ERR "omfs: Invalid superblock (%x)\n",
				   omfs_sb->s_magic);
		goto out_brelse_bh;
	}
	sb->s_magic = OMFS_MAGIC;

	sbi->s_num_blocks = be64_to_cpu(omfs_sb->s_num_blocks);
	sbi->s_blocksize = be32_to_cpu(omfs_sb->s_blocksize);
	sbi->s_mirrors = be32_to_cpu(omfs_sb->s_mirrors);
	sbi->s_root_ino = be64_to_cpu(omfs_sb->s_root_block);
	sbi->s_sys_blocksize = be32_to_cpu(omfs_sb->s_sys_blocksize);
	mutex_init(&sbi->s_bitmap_lock);

	if (sbi->s_sys_blocksize > PAGE_SIZE) {
		printk(KERN_ERR "omfs: sysblock size (%d) is out of range\n",
			sbi->s_sys_blocksize);
		goto out_brelse_bh;
	}

	if (sbi->s_blocksize < sbi->s_sys_blocksize ||
	    sbi->s_blocksize > OMFS_MAX_BLOCK_SIZE) {
		printk(KERN_ERR "omfs: block size (%d) is out of range\n",
			sbi->s_blocksize);
		goto out_brelse_bh;
	}

	/*
	 * Use sys_blocksize as the fs block since it is smaller than a
	 * page while the fs blocksize can be larger.
	 */
	sb_set_blocksize(sb, sbi->s_sys_blocksize);

	/*
	 * ...and the difference goes into a shift.  sys_blocksize is always
	 * a power of two factor of blocksize.
	 */
	sbi->s_block_shift = get_bitmask_order(sbi->s_blocksize) -
		get_bitmask_order(sbi->s_sys_blocksize);

	start = clus_to_blk(sbi, be64_to_cpu(omfs_sb->s_root_block));
	bh2 = sb_bread(sb, start);
	if (!bh2)
		goto out_brelse_bh;

	omfs_rb = (struct omfs_root_block *)bh2->b_data;

	sbi->s_bitmap_ino = be64_to_cpu(omfs_rb->r_bitmap);
	sbi->s_clustersize = be32_to_cpu(omfs_rb->r_clustersize);

	if (sbi->s_num_blocks != be64_to_cpu(omfs_rb->r_num_blocks)) {
		printk(KERN_ERR "omfs: block count discrepancy between "
			"super and root blocks (%llx, %llx)\n",
			(unsigned long long)sbi->s_num_blocks,
			(unsigned long long)be64_to_cpu(omfs_rb->r_num_blocks));
		goto out_brelse_bh2;
	}

	ret = omfs_get_imap(sb);
	if (ret)
		goto out_brelse_bh2;

	sb->s_op = &omfs_sops;

	root = omfs_iget(sb, be64_to_cpu(omfs_rb->r_root_dir));
	if (IS_ERR(root)) {
		ret = PTR_ERR(root);
		goto out_brelse_bh2;
	}

	sb->s_root = d_alloc_root(root);
	if (!sb->s_root) {
		iput(root);
		goto out_brelse_bh2;
	}
	printk(KERN_DEBUG "omfs: Mounted volume %s\n", omfs_rb->r_name);

	ret = 0;
out_brelse_bh2:
	brelse(bh2);
out_brelse_bh:
	brelse(bh);
end:
	return ret;
}

static int omfs_get_sb(struct file_system_type *fs_type,
			int flags, const char *dev_name,
			void *data, struct vfsmount *m)
{
	return get_sb_bdev(fs_type, flags, dev_name, data, omfs_fill_super, m);
}

static struct file_system_type omfs_fs_type = {
	.owner = THIS_MODULE,
	.name = "omfs",
	.get_sb = omfs_get_sb,
	.kill_sb = kill_block_super,
	.fs_flags = FS_REQUIRES_DEV,
};

static int __init init_omfs_fs(void)
{
	return register_filesystem(&omfs_fs_type);
}

static void __exit exit_omfs_fs(void)
{
	unregister_filesystem(&omfs_fs_type);
}

module_init(init_omfs_fs);
module_exit(exit_omfs_fs);
Commit	Line	Data
555e3775 BC	1	/*
	2	* Optimized MPEG FS - inode and super operations.
	3	* Copyright (C) 2006 Bob Copeland <me@bobcopeland.com>
	4	* Released under GPL v2.
	5	*/
555e3775 BC	6	#include <linux/module.h>
555e3775 BC	7	#include <linux/sched.h>
5a0e3ad6	8	#include <linux/slab.h>
555e3775 BC	9	#include <linux/fs.h>
	10	#include <linux/vfs.h>
	11	#include <linux/parser.h>
	12	#include <linux/buffer_head.h>
	13	#include <linux/vmalloc.h>
a9185b41	14	#include <linux/writeback.h>
555e3775 BC	15	#include <linux/crc-itu-t.h>
	16	#include "omfs.h"
	17
	18	MODULE_AUTHOR("Bob Copeland <me@bobcopeland.com>");
	19	MODULE_DESCRIPTION("OMFS (ReplayTV/Karma) Filesystem for Linux");
	20	MODULE_LICENSE("GPL");
	21
	22	struct inode omfs_new_inode(struct inode dir, int mode)
	23	{
	24	struct inode *inode;
	25	u64 new_block;
	26	int err;
	27	int len;
	28	struct omfs_sb_info *sbi = OMFS_SB(dir->i_sb);
	29
	30	inode = new_inode(dir->i_sb);
	31	if (!inode)
	32	return ERR_PTR(-ENOMEM);
	33
	34	err = omfs_allocate_range(dir->i_sb, sbi->s_mirrors, sbi->s_mirrors,
	35	&new_block, &len);
	36	if (err)
	37	goto fail;
	38
	39	inode->i_ino = new_block;
6a9e652c	40	inode_init_owner(inode, NULL, mode);
555e3775 BC	41	inode->i_mapping->a_ops = &omfs_aops;
	42
	43	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
	44	switch (mode & S_IFMT) {
	45	case S_IFDIR:
	46	inode->i_op = &omfs_dir_inops;
	47	inode->i_fop = &omfs_dir_operations;
	48	inode->i_size = sbi->s_sys_blocksize;
	49	inc_nlink(inode);
	50	break;
	51	case S_IFREG:
	52	inode->i_op = &omfs_file_inops;
	53	inode->i_fop = &omfs_file_operations;
	54	inode->i_size = 0;
	55	break;
	56	}
	57
	58	insert_inode_hash(inode);
	59	mark_inode_dirty(inode);
	60	return inode;
	61	fail:
	62	make_bad_inode(inode);
	63	iput(inode);
	64	return ERR_PTR(err);
	65	}
	66
	67	/*
	68	* Update the header checksums for a dirty inode based on its contents.
	69	* Caller is expected to hold the buffer head underlying oi and mark it
	70	* dirty.
	71	*/
	72	static void omfs_update_checksums(struct omfs_inode *oi)
	73	{
	74	int xor, i, ofs = 0, count;
	75	u16 crc = 0;
	76	unsigned char ptr = (unsigned char ) oi;
	77
	78	count = be32_to_cpu(oi->i_head.h_body_size);
	79	ofs = sizeof(struct omfs_header);
	80
	81	crc = crc_itu_t(crc, ptr + ofs, count);
	82	oi->i_head.h_crc = cpu_to_be16(crc);
	83
	84	xor = ptr[0];
	85	for (i = 1; i < OMFS_XOR_COUNT; i++)
	86	xor ^= ptr[i];
	87
	88	oi->i_head.h_check_xor = xor;
	89	}
	90
a9185b41	91	static int __omfs_write_inode(struct inode *inode, int wait)
555e3775 BC	92	{
	93	struct omfs_inode *oi;
	94	struct omfs_sb_info *sbi = OMFS_SB(inode->i_sb);
	95	struct buffer_head bh, bh2;
	96	unsigned int block;
	97	u64 ctime;
	98	int i;
	99	int ret = -EIO;
	100	int sync_failed = 0;
	101
	102	/* get current inode since we may have written sibling ptrs etc. */
	103	block = clus_to_blk(sbi, inode->i_ino);
	104	bh = sb_bread(inode->i_sb, block);
	105	if (!bh)
	106	goto out;
	107
	108	oi = (struct omfs_inode *) bh->b_data;
	109
	110	oi->i_head.h_self = cpu_to_be64(inode->i_ino);
	111	if (S_ISDIR(inode->i_mode))
	112	oi->i_type = OMFS_DIR;
	113	else if (S_ISREG(inode->i_mode))
	114	oi->i_type = OMFS_FILE;
	115	else {
	116	printk(KERN_WARNING "omfs: unknown file type: %d\n",
	117	inode->i_mode);
	118	goto out_brelse;
	119	}
	120
	121	oi->i_head.h_body_size = cpu_to_be32(sbi->s_sys_blocksize -
	122	sizeof(struct omfs_header));
	123	oi->i_head.h_version = 1;
	124	oi->i_head.h_type = OMFS_INODE_NORMAL;
	125	oi->i_head.h_magic = OMFS_IMAGIC;
	126	oi->i_size = cpu_to_be64(inode->i_size);
	127
	128	ctime = inode->i_ctime.tv_sec * 1000LL +
	129	((inode->i_ctime.tv_nsec + 999)/1000);
	130	oi->i_ctime = cpu_to_be64(ctime);
	131
	132	omfs_update_checksums(oi);
	133
	134	mark_buffer_dirty(bh);
	135	if (wait) {
	136	sync_dirty_buffer(bh);
	137	if (buffer_req(bh) && !buffer_uptodate(bh))
	138	sync_failed = 1;
	139	}
	140
	141	/* if mirroring writes, copy to next fsblock */
	142	for (i = 1; i < sbi->s_mirrors; i++) {
	143	bh2 = sb_bread(inode->i_sb, block + i *
	144	(sbi->s_blocksize / sbi->s_sys_blocksize));
	145	if (!bh2)
	146	goto out_brelse;
	147
	148	memcpy(bh2->b_data, bh->b_data, bh->b_size);
	149	mark_buffer_dirty(bh2);
	150	if (wait) {
	151	sync_dirty_buffer(bh2);
	152	if (buffer_req(bh2) && !buffer_uptodate(bh2))
	153	sync_failed = 1;
	154	}
	155	brelse(bh2);
156	}
157	ret = (sync_failed) ? -EIO : 0;
158	out_brelse:
159	brelse(bh);
160	out:
161	return ret;
162	}
163
a9185b41 CH	164	static int omfs_write_inode(struct inode inode, struct writeback_control wbc)
	165	{
	166	return __omfs_write_inode(inode, wbc->sync_mode == WB_SYNC_ALL);
	167	}
	168
555e3775 BC	169	int omfs_sync_inode(struct inode *inode)
555e3775 BC	170	{
a9185b41	171	return __omfs_write_inode(inode, 1);
555e3775 BC	172	}
	173
	174	/*
	175	* called when an entry is deleted, need to clear the bits in the
	176	* bitmaps.
	177	*/
	178	static void omfs_delete_inode(struct inode *inode)
	179	{
	180	truncate_inode_pages(&inode->i_data, 0);
	181
	182	if (S_ISREG(inode->i_mode)) {
	183	inode->i_size = 0;
	184	omfs_shrink_inode(inode);
	185	}
	186
	187	omfs_clear_range(inode->i_sb, inode->i_ino, 2);
	188	clear_inode(inode);
	189	}
	190
	191	struct inode omfs_iget(struct super_block sb, ino_t ino)
	192	{
	193	struct omfs_sb_info *sbi = OMFS_SB(sb);
	194	struct omfs_inode *oi;
	195	struct buffer_head *bh;
	196	unsigned int block;
	197	u64 ctime;
	198	unsigned long nsecs;
	199	struct inode *inode;
	200
	201	inode = iget_locked(sb, ino);
	202	if (!inode)
	203	return ERR_PTR(-ENOMEM);
	204	if (!(inode->i_state & I_NEW))
	205	return inode;
	206
	207	block = clus_to_blk(sbi, ino);
	208	bh = sb_bread(inode->i_sb, block);
	209	if (!bh)
	210	goto iget_failed;
	211
	212	oi = (struct omfs_inode *)bh->b_data;
	213
	214	/* check self */
	215	if (ino != be64_to_cpu(oi->i_head.h_self))
	216	goto fail_bh;
	217
	218	inode->i_uid = sbi->s_uid;
	219	inode->i_gid = sbi->s_gid;
	220
	221	ctime = be64_to_cpu(oi->i_ctime);
	222	nsecs = do_div(ctime, 1000) * 1000L;
	223
	224	inode->i_atime.tv_sec = ctime;
	225	inode->i_mtime.tv_sec = ctime;
	226	inode->i_ctime.tv_sec = ctime;
	227	inode->i_atime.tv_nsec = nsecs;
	228	inode->i_mtime.tv_nsec = nsecs;
	229	inode->i_ctime.tv_nsec = nsecs;
	230
	231	inode->i_mapping->a_ops = &omfs_aops;
	232
	233	switch (oi->i_type) {
	234	case OMFS_DIR:
	235	inode->i_mode = S_IFDIR \| (S_IRWXUGO & ~sbi->s_dmask);
236	inode->i_op = &omfs_dir_inops;
237	inode->i_fop = &omfs_dir_operations;
c963343a	238	inode->i_size = sbi->s_sys_blocksize;
555e3775 BC	239	inc_nlink(inode);
	240	break;
	241	case OMFS_FILE:
	242	inode->i_mode = S_IFREG \| (S_IRWXUGO & ~sbi->s_fmask);
	243	inode->i_fop = &omfs_file_operations;
	244	inode->i_size = be64_to_cpu(oi->i_size);
	245	break;
	246	}
	247	brelse(bh);
	248	unlock_new_inode(inode);
	249	return inode;
	250	fail_bh:
	251	brelse(bh);
	252	iget_failed:
	253	iget_failed(inode);
	254	return ERR_PTR(-EIO);
	255	}
	256
	257	static void omfs_put_super(struct super_block *sb)
	258	{
	259	struct omfs_sb_info *sbi = OMFS_SB(sb);
	260	kfree(sbi->s_imap);
	261	kfree(sbi);
	262	sb->s_fs_info = NULL;
	263	}
	264
	265	static int omfs_statfs(struct dentry dentry, struct kstatfs buf)
	266	{
	267	struct super_block *s = dentry->d_sb;
	268	struct omfs_sb_info *sbi = OMFS_SB(s);
197e671e CL	269	u64 id = huge_encode_dev(s->s_bdev->bd_dev);
197e671e CL	270
555e3775 BC	271	buf->f_type = OMFS_MAGIC;
	272	buf->f_bsize = sbi->s_blocksize;
	273	buf->f_blocks = sbi->s_num_blocks;
	274	buf->f_files = sbi->s_num_blocks;
	275	buf->f_namelen = OMFS_NAMELEN;
197e671e CL	276	buf->f_fsid.val[0] = (u32)id;
197e671e CL	277	buf->f_fsid.val[1] = (u32)(id >> 32);
555e3775 BC	278
	279	buf->f_bfree = buf->f_bavail = buf->f_ffree =
	280	omfs_count_free(s);
197e671e	281
555e3775 BC	282	return 0;
	283	}
	284
b87221de	285	static const struct super_operations omfs_sops = {
555e3775 BC	286	.write_inode = omfs_write_inode,
	287	.delete_inode = omfs_delete_inode,
	288	.put_super = omfs_put_super,
	289	.statfs = omfs_statfs,
	290	.show_options = generic_show_options,
	291	};
	292
	293	/*
	294	* For Rio Karma, there is an on-disk free bitmap whose location is
	295	* stored in the root block. For ReplayTV, there is no such free bitmap
	296	* so we have to walk the tree. Both inodes and file data are allocated
	297	* from the same map. This array can be big (300k) so we allocate
	298	* in units of the blocksize.
	299	*/
	300	static int omfs_get_imap(struct super_block *sb)
	301	{
	302	int bitmap_size;
	303	int array_size;
	304	int count;
	305	struct omfs_sb_info *sbi = OMFS_SB(sb);
	306	struct buffer_head *bh;
	307	unsigned long **ptr;
	308	sector_t block;
	309
	310	bitmap_size = DIV_ROUND_UP(sbi->s_num_blocks, 8);
	311	array_size = DIV_ROUND_UP(bitmap_size, sb->s_blocksize);
	312
	313	if (sbi->s_bitmap_ino == ~0ULL)
	314	goto out;
	315
	316	sbi->s_imap_size = array_size;
	317	sbi->s_imap = kzalloc(array_size * sizeof(unsigned long *), GFP_KERNEL);
	318	if (!sbi->s_imap)
	319	goto nomem;
	320
	321	block = clus_to_blk(sbi, sbi->s_bitmap_ino);
	322	ptr = sbi->s_imap;
	323	for (count = bitmap_size; count > 0; count -= sb->s_blocksize) {
	324	bh = sb_bread(sb, block++);
	325	if (!bh)
	326	goto nomem_free;
	327	*ptr = kmalloc(sb->s_blocksize, GFP_KERNEL);
	328	if (!*ptr) {
	329	brelse(bh);
	330	goto nomem_free;
	331	}
	332	memcpy(*ptr, bh->b_data, sb->s_blocksize);
	333	if (count < sb->s_blocksize)
	334	memset((void )ptr + count, 0xff,
	335	sb->s_blocksize - count);
	336	brelse(bh);
	337	ptr++;
	338	}
	339	out:
	340	return 0;
	341
	342	nomem_free:
	343	for (count = 0; count < array_size; count++)
	344	kfree(sbi->s_imap[count]);
	345
	346	kfree(sbi->s_imap);
	347	nomem:
	348	sbi->s_imap = NULL;
	349	sbi->s_imap_size = 0;
350	return -ENOMEM;
351	}
352
353	enum {
354	Opt_uid, Opt_gid, Opt_umask, Opt_dmask, Opt_fmask
355	};
356
a447c093	357	static const match_table_t tokens = {
555e3775 BC	358	{Opt_uid, "uid=%u"},
	359	{Opt_gid, "gid=%u"},
	360	{Opt_umask, "umask=%o"},
	361	{Opt_dmask, "dmask=%o"},
	362	{Opt_fmask, "fmask=%o"},
	363	};
	364
	365	static int parse_options(char options, struct omfs_sb_info sbi)
	366	{
	367	char *p;
	368	substring_t args[MAX_OPT_ARGS];
	369	int option;
	370
	371	if (!options)
	372	return 1;
	373
	374	while ((p = strsep(&options, ",")) != NULL) {
	375	int token;
	376	if (!*p)
	377	continue;
	378
	379	token = match_token(p, tokens, args);
	380	switch (token) {
	381	case Opt_uid:
	382	if (match_int(&args[0], &option))
	383	return 0;
	384	sbi->s_uid = option;
	385	break;
	386	case Opt_gid:
	387	if (match_int(&args[0], &option))
	388	return 0;
	389	sbi->s_gid = option;
	390	break;
	391	case Opt_umask:
	392	if (match_octal(&args[0], &option))
	393	return 0;
	394	sbi->s_fmask = sbi->s_dmask = option;
	395	break;
	396	case Opt_dmask:
	397	if (match_octal(&args[0], &option))
	398	return 0;
	399	sbi->s_dmask = option;
	400	break;
	401	case Opt_fmask:
	402	if (match_octal(&args[0], &option))
	403	return 0;
	404	sbi->s_fmask = option;
	405	break;
	406	default:
	407	return 0;
	408	}
	409	}
	410	return 1;
	411	}
	412
	413	static int omfs_fill_super(struct super_block sb, void data, int silent)
	414	{
	415	struct buffer_head bh, bh2;
	416	struct omfs_super_block *omfs_sb;
	417	struct omfs_root_block *omfs_rb;
	418	struct omfs_sb_info *sbi;
	419	struct inode *root;
	420	sector_t start;
	421	int ret = -EINVAL;
422
423	save_mount_options(sb, (char *) data);
424
425	sbi = kzalloc(sizeof(struct omfs_sb_info), GFP_KERNEL);
426	if (!sbi)
427	return -ENOMEM;
428
429	sb->s_fs_info = sbi;
430
c222d53e DH	431	sbi->s_uid = current_uid();
c222d53e DH	432	sbi->s_gid = current_gid();
ce3b0f8d	433	sbi->s_dmask = sbi->s_fmask = current_umask();
555e3775 BC	434
	435	if (!parse_options((char *) data, sbi))
	436	goto end;
	437
	438	sb->s_maxbytes = 0xffffffff;
	439
	440	sb_set_blocksize(sb, 0x200);
	441
	442	bh = sb_bread(sb, 0);
	443	if (!bh)
	444	goto end;
	445
	446	omfs_sb = (struct omfs_super_block *)bh->b_data;
	447
	448	if (omfs_sb->s_magic != cpu_to_be32(OMFS_MAGIC)) {
	449	if (!silent)
	450	printk(KERN_ERR "omfs: Invalid superblock (%x)\n",
	451	omfs_sb->s_magic);
	452	goto out_brelse_bh;
	453	}
	454	sb->s_magic = OMFS_MAGIC;
	455
	456	sbi->s_num_blocks = be64_to_cpu(omfs_sb->s_num_blocks);
	457	sbi->s_blocksize = be32_to_cpu(omfs_sb->s_blocksize);
	458	sbi->s_mirrors = be32_to_cpu(omfs_sb->s_mirrors);
	459	sbi->s_root_ino = be64_to_cpu(omfs_sb->s_root_block);
	460	sbi->s_sys_blocksize = be32_to_cpu(omfs_sb->s_sys_blocksize);
	461	mutex_init(&sbi->s_bitmap_lock);
	462
	463	if (sbi->s_sys_blocksize > PAGE_SIZE) {
	464	printk(KERN_ERR "omfs: sysblock size (%d) is out of range\n",
	465	sbi->s_sys_blocksize);
	466	goto out_brelse_bh;
	467	}
	468
	469	if (sbi->s_blocksize < sbi->s_sys_blocksize \|\|
	470	sbi->s_blocksize > OMFS_MAX_BLOCK_SIZE) {
	471	printk(KERN_ERR "omfs: block size (%d) is out of range\n",
	472	sbi->s_blocksize);
	473	goto out_brelse_bh;
	474	}
	475
	476	/*
	477	* Use sys_blocksize as the fs block since it is smaller than a
	478	* page while the fs blocksize can be larger.
	479	*/
	480	sb_set_blocksize(sb, sbi->s_sys_blocksize);
	481
	482	/*
	483	* ...and the difference goes into a shift. sys_blocksize is always
	484	* a power of two factor of blocksize.
	485	*/
	486	sbi->s_block_shift = get_bitmask_order(sbi->s_blocksize) -
	487	get_bitmask_order(sbi->s_sys_blocksize);
	488
	489	start = clus_to_blk(sbi, be64_to_cpu(omfs_sb->s_root_block));
	490	bh2 = sb_bread(sb, start);
	491	if (!bh2)
	492	goto out_brelse_bh;
	493
	494	omfs_rb = (struct omfs_root_block *)bh2->b_data;
	495
	496	sbi->s_bitmap_ino = be64_to_cpu(omfs_rb->r_bitmap);
	497	sbi->s_clustersize = be32_to_cpu(omfs_rb->r_clustersize);
498
499	if (sbi->s_num_blocks != be64_to_cpu(omfs_rb->r_num_blocks)) {
500	printk(KERN_ERR "omfs: block count discrepancy between "
501	"super and root blocks (%llx, %llx)\n",
dc60bf1d AB	502	(unsigned long long)sbi->s_num_blocks,
dc60bf1d AB	503	(unsigned long long)be64_to_cpu(omfs_rb->r_num_blocks));
555e3775 BC	504	goto out_brelse_bh2;
	505	}
	506
	507	ret = omfs_get_imap(sb);
	508	if (ret)
	509	goto out_brelse_bh2;
	510
	511	sb->s_op = &omfs_sops;
	512
	513	root = omfs_iget(sb, be64_to_cpu(omfs_rb->r_root_dir));
	514	if (IS_ERR(root)) {
	515	ret = PTR_ERR(root);
	516	goto out_brelse_bh2;
	517	}
	518
	519	sb->s_root = d_alloc_root(root);
	520	if (!sb->s_root) {
	521	iput(root);
	522	goto out_brelse_bh2;
	523	}
	524	printk(KERN_DEBUG "omfs: Mounted volume %s\n", omfs_rb->r_name);
	525
	526	ret = 0;
	527	out_brelse_bh2:
	528	brelse(bh2);
	529	out_brelse_bh:
	530	brelse(bh);
	531	end:
	532	return ret;
	533	}
	534
	535	static int omfs_get_sb(struct file_system_type *fs_type,
	536	int flags, const char *dev_name,
	537	void data, struct vfsmount m)
	538	{
	539	return get_sb_bdev(fs_type, flags, dev_name, data, omfs_fill_super, m);
	540	}
	541
	542	static struct file_system_type omfs_fs_type = {
	543	.owner = THIS_MODULE,
	544	.name = "omfs",
	545	.get_sb = omfs_get_sb,
	546	.kill_sb = kill_block_super,
	547	.fs_flags = FS_REQUIRES_DEV,
	548	};
	549
	550	static int __init init_omfs_fs(void)
	551	{
	552	return register_filesystem(&omfs_fs_type);
	553	}
	554
	555	static void __exit exit_omfs_fs(void)
	556	{
	557	unregister_filesystem(&omfs_fs_type);
	558	}
	559
	560	module_init(init_omfs_fs);
	561	module_exit(exit_omfs_fs);