[net-next-2.6.git] / fs / logfs / super.c

/*
 * fs/logfs/super.c
 *
 * As should be obvious for Linux kernel code, license is GPLv2
 *
 * Copyright (c) 2005-2008 Joern Engel <joern@logfs.org>
 *
 * Generally contains mount/umount code and also serves as a dump area for
 * any functions that don't fit elsewhere and neither justify a file of their
 * own.
 */
#include "logfs.h"
#include <linux/bio.h>
#include <linux/slab.h>
#include <linux/blkdev.h>
#include <linux/mtd/mtd.h>
#include <linux/statfs.h>
#include <linux/buffer_head.h>

static DEFINE_MUTEX(emergency_mutex);
static struct page *emergency_page;

struct page *emergency_read_begin(struct address_space *mapping, pgoff_t index)
{
	filler_t *filler = (filler_t *)mapping->a_ops->readpage;
	struct page *page;
	int err;

	page = read_cache_page(mapping, index, filler, NULL);
	if (page)
		return page;

	/* No more pages available, switch to emergency page */
	printk(KERN_INFO"Logfs: Using emergency page\n");
	mutex_lock(&emergency_mutex);
	err = filler(NULL, emergency_page);
	if (err) {
		mutex_unlock(&emergency_mutex);
		printk(KERN_EMERG"Logfs: Error reading emergency page\n");
		return ERR_PTR(err);
	}
	return emergency_page;
}

void emergency_read_end(struct page *page)
{
	if (page == emergency_page)
		mutex_unlock(&emergency_mutex);
	else
		page_cache_release(page);
}

static void dump_segfile(struct super_block *sb)
{
	struct logfs_super *super = logfs_super(sb);
	struct logfs_segment_entry se;
	u32 segno;

	for (segno = 0; segno < super->s_no_segs; segno++) {
		logfs_get_segment_entry(sb, segno, &se);
		printk("%3x: %6x %8x", segno, be32_to_cpu(se.ec_level),
				be32_to_cpu(se.valid));
		if (++segno < super->s_no_segs) {
			logfs_get_segment_entry(sb, segno, &se);
			printk(" %6x %8x", be32_to_cpu(se.ec_level),
					be32_to_cpu(se.valid));
		}
		if (++segno < super->s_no_segs) {
			logfs_get_segment_entry(sb, segno, &se);
			printk(" %6x %8x", be32_to_cpu(se.ec_level),
					be32_to_cpu(se.valid));
		}
		if (++segno < super->s_no_segs) {
			logfs_get_segment_entry(sb, segno, &se);
			printk(" %6x %8x", be32_to_cpu(se.ec_level),
					be32_to_cpu(se.valid));
		}
		printk("\n");
	}
}

/*
 * logfs_crash_dump - dump debug information to device
 *
 * The LogFS superblock only occupies part of a segment.  This function will
 * write as much debug information as it can gather into the spare space.
 */
void logfs_crash_dump(struct super_block *sb)
{
	dump_segfile(sb);
}

/*
 * TODO: move to lib/string.c
 */
/**
 * memchr_inv - Find a character in an area of memory.
 * @s: The memory area
 * @c: The byte to search for
 * @n: The size of the area.
 *
 * returns the address of the first character other than @c, or %NULL
 * if the whole buffer contains just @c.
 */
void *memchr_inv(const void *s, int c, size_t n)
{
	const unsigned char *p = s;
	while (n-- != 0)
		if ((unsigned char)c != *p++)
			return (void *)(p - 1);

	return NULL;
}

/*
 * FIXME: There should be a reserve for root, similar to ext2.
 */
int logfs_statfs(struct dentry *dentry, struct kstatfs *stats)
{
	struct super_block *sb = dentry->d_sb;
	struct logfs_super *super = logfs_super(sb);

	stats->f_type		= LOGFS_MAGIC_U32;
	stats->f_bsize		= sb->s_blocksize;
	stats->f_blocks		= super->s_size >> LOGFS_BLOCK_BITS >> 3;
	stats->f_bfree		= super->s_free_bytes >> sb->s_blocksize_bits;
	stats->f_bavail		= super->s_free_bytes >> sb->s_blocksize_bits;
	stats->f_files		= 0;
	stats->f_ffree		= 0;
	stats->f_namelen	= LOGFS_MAX_NAMELEN;
	return 0;
}

static int logfs_sb_set(struct super_block *sb, void *_super)
{
	struct logfs_super *super = _super;

	sb->s_fs_info = super;
	sb->s_mtd = super->s_mtd;
	sb->s_bdev = super->s_bdev;
	if (sb->s_bdev)
		sb->s_bdi = &bdev_get_queue(sb->s_bdev)->backing_dev_info;
	if (sb->s_mtd)
		sb->s_bdi = sb->s_mtd->backing_dev_info;
	return 0;
}

static int logfs_sb_test(struct super_block *sb, void *_super)
{
	struct logfs_super *super = _super;
	struct mtd_info *mtd = super->s_mtd;

	if (mtd && sb->s_mtd == mtd)
		return 1;
	if (super->s_bdev && sb->s_bdev == super->s_bdev)
		return 1;
	return 0;
}

static void set_segment_header(struct logfs_segment_header *sh, u8 type,
		u8 level, u32 segno, u32 ec)
{
	sh->pad = 0;
	sh->type = type;
	sh->level = level;
	sh->segno = cpu_to_be32(segno);
	sh->ec = cpu_to_be32(ec);
	sh->gec = cpu_to_be64(segno);
	sh->crc = logfs_crc32(sh, LOGFS_SEGMENT_HEADERSIZE, 4);
}

static void logfs_write_ds(struct super_block *sb, struct logfs_disk_super *ds,
		u32 segno, u32 ec)
{
	struct logfs_super *super = logfs_super(sb);
	struct logfs_segment_header *sh = &ds->ds_sh;
	int i;

	memset(ds, 0, sizeof(*ds));
	set_segment_header(sh, SEG_SUPER, 0, segno, ec);

	ds->ds_ifile_levels	= super->s_ifile_levels;
	ds->ds_iblock_levels	= super->s_iblock_levels;
	ds->ds_data_levels	= super->s_data_levels; /* XXX: Remove */
	ds->ds_segment_shift	= super->s_segshift;
	ds->ds_block_shift	= sb->s_blocksize_bits;
	ds->ds_write_shift	= super->s_writeshift;
	ds->ds_filesystem_size	= cpu_to_be64(super->s_size);
	ds->ds_segment_size	= cpu_to_be32(super->s_segsize);
	ds->ds_bad_seg_reserve	= cpu_to_be32(super->s_bad_seg_reserve);
	ds->ds_feature_incompat	= cpu_to_be64(super->s_feature_incompat);
	ds->ds_feature_ro_compat= cpu_to_be64(super->s_feature_ro_compat);
	ds->ds_feature_compat	= cpu_to_be64(super->s_feature_compat);
	ds->ds_feature_flags	= cpu_to_be64(super->s_feature_flags);
	ds->ds_root_reserve	= cpu_to_be64(super->s_root_reserve);
	ds->ds_speed_reserve	= cpu_to_be64(super->s_speed_reserve);
	journal_for_each(i)
		ds->ds_journal_seg[i] = cpu_to_be32(super->s_journal_seg[i]);
	ds->ds_magic		= cpu_to_be64(LOGFS_MAGIC);
	ds->ds_crc = logfs_crc32(ds, sizeof(*ds),
			LOGFS_SEGMENT_HEADERSIZE + 12);
}

static int write_one_sb(struct super_block *sb,
		struct page *(*find_sb)(struct super_block *sb, u64 *ofs))
{
	struct logfs_super *super = logfs_super(sb);
	struct logfs_disk_super *ds;
	struct logfs_segment_entry se;
	struct page *page;
	u64 ofs;
	u32 ec, segno;
	int err;

	page = find_sb(sb, &ofs);
	if (!page)
		return -EIO;
	ds = page_address(page);
	segno = seg_no(sb, ofs);
	logfs_get_segment_entry(sb, segno, &se);
	ec = be32_to_cpu(se.ec_level) >> 4;
	ec++;
	logfs_set_segment_erased(sb, segno, ec, 0);
	logfs_write_ds(sb, ds, segno, ec);
	err = super->s_devops->write_sb(sb, page);
	page_cache_release(page);
	return err;
}

int logfs_write_sb(struct super_block *sb)
{
	struct logfs_super *super = logfs_super(sb);
	int err;

	/* First superblock */
	err = write_one_sb(sb, super->s_devops->find_first_sb);
	if (err)
		return err;

	/* Last superblock */
	err = write_one_sb(sb, super->s_devops->find_last_sb);
	if (err)
		return err;
	return 0;
}

static int ds_cmp(const void *ds0, const void *ds1)
{
	size_t len = sizeof(struct logfs_disk_super);

	/* We know the segment headers differ, so ignore them */
	len -= LOGFS_SEGMENT_HEADERSIZE;
	ds0 += LOGFS_SEGMENT_HEADERSIZE;
	ds1 += LOGFS_SEGMENT_HEADERSIZE;
	return memcmp(ds0, ds1, len);
}

static int logfs_recover_sb(struct super_block *sb)
{
	struct logfs_super *super = logfs_super(sb);
	struct logfs_disk_super _ds0, *ds0 = &_ds0;
	struct logfs_disk_super _ds1, *ds1 = &_ds1;
	int err, valid0, valid1;

	/* read first superblock */
	err = wbuf_read(sb, super->s_sb_ofs[0], sizeof(*ds0), ds0);
	if (err)
		return err;
	/* read last superblock */
	err = wbuf_read(sb, super->s_sb_ofs[1], sizeof(*ds1), ds1);
	if (err)
		return err;
	valid0 = logfs_check_ds(ds0) == 0;
	valid1 = logfs_check_ds(ds1) == 0;

	if (!valid0 && valid1) {
		printk(KERN_INFO"First superblock is invalid - fixing.\n");
		return write_one_sb(sb, super->s_devops->find_first_sb);
	}
	if (valid0 && !valid1) {
		printk(KERN_INFO"Last superblock is invalid - fixing.\n");
		return write_one_sb(sb, super->s_devops->find_last_sb);
	}
	if (valid0 && valid1 && ds_cmp(ds0, ds1)) {
		printk(KERN_INFO"Superblocks don't match - fixing.\n");
		return logfs_write_sb(sb);
	}
	/* If neither is valid now, something's wrong.  Didn't we properly
	 * check them before?!? */
	BUG_ON(!valid0 && !valid1);
	return 0;
}

static int logfs_make_writeable(struct super_block *sb)
{
	int err;

	err = logfs_open_segfile(sb);
	if (err)
		return err;

	/* Repair any broken superblock copies */
	err = logfs_recover_sb(sb);
	if (err)
		return err;

	/* Check areas for trailing unaccounted data */
	err = logfs_check_areas(sb);
	if (err)
		return err;

	/* Do one GC pass before any data gets dirtied */
	logfs_gc_pass(sb);

	/* after all initializations are done, replay the journal
	 * for rw-mounts, if necessary */
	err = logfs_replay_journal(sb);
	if (err)
		return err;

	return 0;
}

static int logfs_get_sb_final(struct super_block *sb, struct vfsmount *mnt)
{
	struct logfs_super *super = logfs_super(sb);
	struct inode *rootdir;
	int err;

	/* root dir */
	rootdir = logfs_iget(sb, LOGFS_INO_ROOT);
	if (IS_ERR(rootdir))
		goto fail;

	sb->s_root = d_alloc_root(rootdir);
	if (!sb->s_root) {
		iput(rootdir);
		goto fail;
	}

	super->s_erase_page = alloc_pages(GFP_KERNEL, 0);
	if (!super->s_erase_page)
		goto fail;
	memset(page_address(super->s_erase_page), 0xFF, PAGE_SIZE);

	/* FIXME: check for read-only mounts */
	err = logfs_make_writeable(sb);
	if (err)
		goto fail1;

	log_super("LogFS: Finished mounting\n");
	simple_set_mnt(mnt, sb);
	return 0;

fail1:
	__free_page(super->s_erase_page);
fail:
	iput(logfs_super(sb)->s_master_inode);
	return -EIO;
}

int logfs_check_ds(struct logfs_disk_super *ds)
{
	struct logfs_segment_header *sh = &ds->ds_sh;

	if (ds->ds_magic != cpu_to_be64(LOGFS_MAGIC))
		return -EINVAL;
	if (sh->crc != logfs_crc32(sh, LOGFS_SEGMENT_HEADERSIZE, 4))
		return -EINVAL;
	if (ds->ds_crc != logfs_crc32(ds, sizeof(*ds),
				LOGFS_SEGMENT_HEADERSIZE + 12))
		return -EINVAL;
	return 0;
}

static struct page *find_super_block(struct super_block *sb)
{
	struct logfs_super *super = logfs_super(sb);
	struct page *first, *last;

	first = super->s_devops->find_first_sb(sb, &super->s_sb_ofs[0]);
	if (!first || IS_ERR(first))
		return NULL;
	last = super->s_devops->find_last_sb(sb, &super->s_sb_ofs[1]);
	if (!last || IS_ERR(first)) {
		page_cache_release(first);
		return NULL;
	}

	if (!logfs_check_ds(page_address(first))) {
		page_cache_release(last);
		return first;
	}

	/* First one didn't work, try the second superblock */
	if (!logfs_check_ds(page_address(last))) {
		page_cache_release(first);
		return last;
	}

	/* Neither worked, sorry folks */
	page_cache_release(first);
	page_cache_release(last);
	return NULL;
}

static int __logfs_read_sb(struct super_block *sb)
{
	struct logfs_super *super = logfs_super(sb);
	struct page *page;
	struct logfs_disk_super *ds;
	int i;

	page = find_super_block(sb);
	if (!page)
		return -EIO;

	ds = page_address(page);
	super->s_size = be64_to_cpu(ds->ds_filesystem_size);
	super->s_root_reserve = be64_to_cpu(ds->ds_root_reserve);
	super->s_speed_reserve = be64_to_cpu(ds->ds_speed_reserve);
	super->s_bad_seg_reserve = be32_to_cpu(ds->ds_bad_seg_reserve);
	super->s_segsize = 1 << ds->ds_segment_shift;
	super->s_segmask = (1 << ds->ds_segment_shift) - 1;
	super->s_segshift = ds->ds_segment_shift;
	sb->s_blocksize = 1 << ds->ds_block_shift;
	sb->s_blocksize_bits = ds->ds_block_shift;
	super->s_writesize = 1 << ds->ds_write_shift;
	super->s_writeshift = ds->ds_write_shift;
	super->s_no_segs = super->s_size >> super->s_segshift;
	super->s_no_blocks = super->s_segsize >> sb->s_blocksize_bits;
	super->s_feature_incompat = be64_to_cpu(ds->ds_feature_incompat);
	super->s_feature_ro_compat = be64_to_cpu(ds->ds_feature_ro_compat);
	super->s_feature_compat = be64_to_cpu(ds->ds_feature_compat);
	super->s_feature_flags = be64_to_cpu(ds->ds_feature_flags);

	journal_for_each(i)
		super->s_journal_seg[i] = be32_to_cpu(ds->ds_journal_seg[i]);

	super->s_ifile_levels = ds->ds_ifile_levels;
	super->s_iblock_levels = ds->ds_iblock_levels;
	super->s_data_levels = ds->ds_data_levels;
	super->s_total_levels = super->s_ifile_levels + super->s_iblock_levels
		+ super->s_data_levels;
	page_cache_release(page);
	return 0;
}

static int logfs_read_sb(struct super_block *sb, int read_only)
{
	struct logfs_super *super = logfs_super(sb);
	int ret;

	super->s_btree_pool = mempool_create(32, btree_alloc, btree_free, NULL);
	if (!super->s_btree_pool)
		return -ENOMEM;

	btree_init_mempool64(&super->s_shadow_tree.new, super->s_btree_pool);
	btree_init_mempool64(&super->s_shadow_tree.old, super->s_btree_pool);
	btree_init_mempool32(&super->s_shadow_tree.segment_map,
			super->s_btree_pool);

	ret = logfs_init_mapping(sb);
	if (ret)
		return ret;

	ret = __logfs_read_sb(sb);
	if (ret)
		return ret;

	if (super->s_feature_incompat & ~LOGFS_FEATURES_INCOMPAT)
		return -EIO;
	if ((super->s_feature_ro_compat & ~LOGFS_FEATURES_RO_COMPAT) &&
			!read_only)
		return -EIO;

	mutex_init(&super->s_dirop_mutex);
	mutex_init(&super->s_object_alias_mutex);
	INIT_LIST_HEAD(&super->s_freeing_list);

	ret = logfs_init_rw(sb);
	if (ret)
		return ret;

	ret = logfs_init_areas(sb);
	if (ret)
		return ret;

	ret = logfs_init_gc(sb);
	if (ret)
		return ret;

	ret = logfs_init_journal(sb);
	if (ret)
		return ret;

	return 0;
}

static void logfs_kill_sb(struct super_block *sb)
{
	struct logfs_super *super = logfs_super(sb);

	log_super("LogFS: Start unmounting\n");
	/* Alias entries slow down mount, so evict as many as possible */
	sync_filesystem(sb);
	logfs_write_anchor(sb);

	/*
	 * From this point on alias entries are simply dropped - and any
	 * writes to the object store are considered bugs.
	 */
	super->s_flags |= LOGFS_SB_FLAG_SHUTDOWN;
	log_super("LogFS: Now in shutdown\n");
	generic_shutdown_super(sb);

	BUG_ON(super->s_dirty_used_bytes || super->s_dirty_free_bytes);

	logfs_cleanup_gc(sb);
	logfs_cleanup_journal(sb);
	logfs_cleanup_areas(sb);
	logfs_cleanup_rw(sb);
	if (super->s_erase_page)
		__free_page(super->s_erase_page);
	super->s_devops->put_device(sb);
	logfs_mempool_destroy(super->s_btree_pool);
	logfs_mempool_destroy(super->s_alias_pool);
	kfree(super);
	log_super("LogFS: Finished unmounting\n");
}

int logfs_get_sb_device(struct file_system_type *type, int flags,
		struct mtd_info *mtd, struct block_device *bdev,
		const struct logfs_device_ops *devops, struct vfsmount *mnt)
{
	struct logfs_super *super;
	struct super_block *sb;
	int err = -ENOMEM;
	static int mount_count;

	log_super("LogFS: Start mount %x\n", mount_count++);
	super = kzalloc(sizeof(*super), GFP_KERNEL);
	if (!super)
		goto err0;

	super->s_mtd	= mtd;
	super->s_bdev	= bdev;
	err = -EINVAL;
	sb = sget(type, logfs_sb_test, logfs_sb_set, super);
	if (IS_ERR(sb))
		goto err0;

	if (sb->s_root) {
		/* Device is already in use */
		err = 0;
		simple_set_mnt(mnt, sb);
		goto err0;
	}

	super->s_devops = devops;

	/*
	 * sb->s_maxbytes is limited to 8TB.  On 32bit systems, the page cache
	 * only covers 16TB and the upper 8TB are used for indirect blocks.
	 * On 64bit system we could bump up the limit, but that would make
	 * the filesystem incompatible with 32bit systems.
	 */
	sb->s_maxbytes	= (1ull << 43) - 1;
	sb->s_op	= &logfs_super_operations;
	sb->s_flags	= flags | MS_NOATIME;

	err = logfs_read_sb(sb, sb->s_flags & MS_RDONLY);
	if (err)
		goto err1;

	sb->s_flags |= MS_ACTIVE;
	err = logfs_get_sb_final(sb, mnt);
	if (err)
		goto err1;
	return 0;

err1:
	deactivate_locked_super(sb);
	return err;
err0:
	kfree(super);
	//devops->put_device(sb);
	return err;
}

static int logfs_get_sb(struct file_system_type *type, int flags,
		const char *devname, void *data, struct vfsmount *mnt)
{
	ulong mtdnr;

	if (!devname)
		return logfs_get_sb_bdev(type, flags, devname, mnt);
	if (strncmp(devname, "mtd", 3))
		return logfs_get_sb_bdev(type, flags, devname, mnt);

	{
		char *garbage;
		mtdnr = simple_strtoul(devname+3, &garbage, 0);
		if (*garbage)
			return -EINVAL;
	}

	return logfs_get_sb_mtd(type, flags, mtdnr, mnt);
}

static struct file_system_type logfs_fs_type = {
	.owner		= THIS_MODULE,
	.name		= "logfs",
	.get_sb		= logfs_get_sb,
	.kill_sb	= logfs_kill_sb,
	.fs_flags	= FS_REQUIRES_DEV,

};

static int __init logfs_init(void)
{
	int ret;

	emergency_page = alloc_pages(GFP_KERNEL, 0);
	if (!emergency_page)
		return -ENOMEM;

	ret = logfs_compr_init();
	if (ret)
		goto out1;

	ret = logfs_init_inode_cache();
	if (ret)
		goto out2;

	return register_filesystem(&logfs_fs_type);
out2:
	logfs_compr_exit();
out1:
	__free_pages(emergency_page, 0);
	return ret;
}

static void __exit logfs_exit(void)
{
	unregister_filesystem(&logfs_fs_type);
	logfs_destroy_inode_cache();
	logfs_compr_exit();
	__free_pages(emergency_page, 0);
}

module_init(logfs_init);
module_exit(logfs_exit);

MODULE_LICENSE("GPL v2");
MODULE_AUTHOR("Joern Engel <joern@logfs.org>");
MODULE_DESCRIPTION("scalable flash filesystem");
Commit	Line	Data
5db53f3e JE	1	/*
	2	* fs/logfs/super.c
	3	*
	4	* As should be obvious for Linux kernel code, license is GPLv2
	5	*
	6	* Copyright (c) 2005-2008 Joern Engel <joern@logfs.org>
	7	*
	8	* Generally contains mount/umount code and also serves as a dump area for
	9	* any functions that don't fit elsewhere and neither justify a file of their
	10	* own.
	11	*/
	12	#include "logfs.h"
	13	#include <linux/bio.h>
5a0e3ad6	14	#include <linux/slab.h>
b8639077	15	#include <linux/blkdev.h>
5db53f3e JE	16	#include <linux/mtd/mtd.h>
	17	#include <linux/statfs.h>
	18	#include <linux/buffer_head.h>
	19
	20	static DEFINE_MUTEX(emergency_mutex);
	21	static struct page *emergency_page;
	22
	23	struct page emergency_read_begin(struct address_space mapping, pgoff_t index)
	24	{
	25	filler_t filler = (filler_t )mapping->a_ops->readpage;
	26	struct page *page;
	27	int err;
	28
	29	page = read_cache_page(mapping, index, filler, NULL);
	30	if (page)
	31	return page;
	32
	33	/* No more pages available, switch to emergency page */
	34	printk(KERN_INFO"Logfs: Using emergency page\n");
	35	mutex_lock(&emergency_mutex);
	36	err = filler(NULL, emergency_page);
	37	if (err) {
	38	mutex_unlock(&emergency_mutex);
	39	printk(KERN_EMERG"Logfs: Error reading emergency page\n");
	40	return ERR_PTR(err);
	41	}
	42	return emergency_page;
	43	}
	44
	45	void emergency_read_end(struct page *page)
	46	{
	47	if (page == emergency_page)
	48	mutex_unlock(&emergency_mutex);
	49	else
	50	page_cache_release(page);
	51	}
	52
	53	static void dump_segfile(struct super_block *sb)
	54	{
	55	struct logfs_super *super = logfs_super(sb);
	56	struct logfs_segment_entry se;
	57	u32 segno;
	58
	59	for (segno = 0; segno < super->s_no_segs; segno++) {
	60	logfs_get_segment_entry(sb, segno, &se);
	61	printk("%3x: %6x %8x", segno, be32_to_cpu(se.ec_level),
	62	be32_to_cpu(se.valid));
	63	if (++segno < super->s_no_segs) {
	64	logfs_get_segment_entry(sb, segno, &se);
	65	printk(" %6x %8x", be32_to_cpu(se.ec_level),
	66	be32_to_cpu(se.valid));
	67	}
	68	if (++segno < super->s_no_segs) {
	69	logfs_get_segment_entry(sb, segno, &se);
	70	printk(" %6x %8x", be32_to_cpu(se.ec_level),
	71	be32_to_cpu(se.valid));
	72	}
	73	if (++segno < super->s_no_segs) {
	74	logfs_get_segment_entry(sb, segno, &se);
	75	printk(" %6x %8x", be32_to_cpu(se.ec_level),
	76	be32_to_cpu(se.valid));
	77	}
	78	printk("\n");
	79	}
80	}
81
82	/*
83	* logfs_crash_dump - dump debug information to device
84	*
85	* The LogFS superblock only occupies part of a segment. This function will
86	* write as much debug information as it can gather into the spare space.
87	*/
88	void logfs_crash_dump(struct super_block *sb)
89	{
90	dump_segfile(sb);
91	}
92
93	/*
94	* TODO: move to lib/string.c
95	*/
96	/**
97	* memchr_inv - Find a character in an area of memory.
98	* @s: The memory area
99	* @c: The byte to search for
100	* @n: The size of the area.
101	*
102	* returns the address of the first character other than @c, or %NULL
103	* if the whole buffer contains just @c.
104	*/
105	void memchr_inv(const void s, int c, size_t n)
106	{
107	const unsigned char *p = s;
108	while (n-- != 0)
109	if ((unsigned char)c != *p++)
110	return (void *)(p - 1);
111
112	return NULL;
113	}
114
115	/*
116	* FIXME: There should be a reserve for root, similar to ext2.
117	*/
118	int logfs_statfs(struct dentry dentry, struct kstatfs stats)
119	{
120	struct super_block *sb = dentry->d_sb;
121	struct logfs_super *super = logfs_super(sb);
122
123	stats->f_type = LOGFS_MAGIC_U32;
124	stats->f_bsize = sb->s_blocksize;
125	stats->f_blocks = super->s_size >> LOGFS_BLOCK_BITS >> 3;
126	stats->f_bfree = super->s_free_bytes >> sb->s_blocksize_bits;
127	stats->f_bavail = super->s_free_bytes >> sb->s_blocksize_bits;
128	stats->f_files = 0;
129	stats->f_ffree = 0;
130	stats->f_namelen = LOGFS_MAX_NAMELEN;
131	return 0;
132	}
133
134	static int logfs_sb_set(struct super_block sb, void _super)
135	{
136	struct logfs_super *super = _super;
137
138	sb->s_fs_info = super;
139	sb->s_mtd = super->s_mtd;
140	sb->s_bdev = super->s_bdev;
b8639077 JE	141	if (sb->s_bdev)
	142	sb->s_bdi = &bdev_get_queue(sb->s_bdev)->backing_dev_info;
	143	if (sb->s_mtd)
	144	sb->s_bdi = sb->s_mtd->backing_dev_info;
5db53f3e JE	145	return 0;
	146	}
	147
	148	static int logfs_sb_test(struct super_block sb, void _super)
	149	{
	150	struct logfs_super *super = _super;
	151	struct mtd_info *mtd = super->s_mtd;
	152
	153	if (mtd && sb->s_mtd == mtd)
	154	return 1;
	155	if (super->s_bdev && sb->s_bdev == super->s_bdev)
	156	return 1;
	157	return 0;
	158	}
	159
	160	static void set_segment_header(struct logfs_segment_header *sh, u8 type,
	161	u8 level, u32 segno, u32 ec)
	162	{
	163	sh->pad = 0;
	164	sh->type = type;
	165	sh->level = level;
	166	sh->segno = cpu_to_be32(segno);
	167	sh->ec = cpu_to_be32(ec);
	168	sh->gec = cpu_to_be64(segno);
	169	sh->crc = logfs_crc32(sh, LOGFS_SEGMENT_HEADERSIZE, 4);
	170	}
	171
	172	static void logfs_write_ds(struct super_block sb, struct logfs_disk_super ds,
	173	u32 segno, u32 ec)
	174	{
	175	struct logfs_super *super = logfs_super(sb);
	176	struct logfs_segment_header *sh = &ds->ds_sh;
	177	int i;
	178
	179	memset(ds, 0, sizeof(*ds));
	180	set_segment_header(sh, SEG_SUPER, 0, segno, ec);
	181
	182	ds->ds_ifile_levels = super->s_ifile_levels;
	183	ds->ds_iblock_levels = super->s_iblock_levels;
	184	ds->ds_data_levels = super->s_data_levels; /* XXX: Remove */
	185	ds->ds_segment_shift = super->s_segshift;
	186	ds->ds_block_shift = sb->s_blocksize_bits;
	187	ds->ds_write_shift = super->s_writeshift;
	188	ds->ds_filesystem_size = cpu_to_be64(super->s_size);
	189	ds->ds_segment_size = cpu_to_be32(super->s_segsize);
	190	ds->ds_bad_seg_reserve = cpu_to_be32(super->s_bad_seg_reserve);
	191	ds->ds_feature_incompat = cpu_to_be64(super->s_feature_incompat);
	192	ds->ds_feature_ro_compat= cpu_to_be64(super->s_feature_ro_compat);
	193	ds->ds_feature_compat = cpu_to_be64(super->s_feature_compat);
	194	ds->ds_feature_flags = cpu_to_be64(super->s_feature_flags);
	195	ds->ds_root_reserve = cpu_to_be64(super->s_root_reserve);
	196	ds->ds_speed_reserve = cpu_to_be64(super->s_speed_reserve);
	197	journal_for_each(i)
	198	ds->ds_journal_seg[i] = cpu_to_be32(super->s_journal_seg[i]);
	199	ds->ds_magic = cpu_to_be64(LOGFS_MAGIC);
	200	ds->ds_crc = logfs_crc32(ds, sizeof(*ds),
	201	LOGFS_SEGMENT_HEADERSIZE + 12);
	202	}
	203
	204	static int write_one_sb(struct super_block *sb,
	205	struct page (find_sb)(struct super_block sb, u64 ofs))
	206	{
	207	struct logfs_super *super = logfs_super(sb);
	208	struct logfs_disk_super *ds;
209	struct logfs_segment_entry se;
210	struct page *page;
211	u64 ofs;
212	u32 ec, segno;
213	int err;
214
215	page = find_sb(sb, &ofs);
216	if (!page)
217	return -EIO;
218	ds = page_address(page);
219	segno = seg_no(sb, ofs);
220	logfs_get_segment_entry(sb, segno, &se);
221	ec = be32_to_cpu(se.ec_level) >> 4;
222	ec++;
223	logfs_set_segment_erased(sb, segno, ec, 0);
224	logfs_write_ds(sb, ds, segno, ec);
225	err = super->s_devops->write_sb(sb, page);
226	page_cache_release(page);
227	return err;
228	}
229
230	int logfs_write_sb(struct super_block *sb)
231	{
232	struct logfs_super *super = logfs_super(sb);
233	int err;
234
235	/* First superblock */
236	err = write_one_sb(sb, super->s_devops->find_first_sb);
237	if (err)
238	return err;
239
240	/* Last superblock */
241	err = write_one_sb(sb, super->s_devops->find_last_sb);
242	if (err)
243	return err;
244	return 0;
245	}
246
247	static int ds_cmp(const void ds0, const void ds1)
248	{
249	size_t len = sizeof(struct logfs_disk_super);
250
251	/* We know the segment headers differ, so ignore them */
252	len -= LOGFS_SEGMENT_HEADERSIZE;
253	ds0 += LOGFS_SEGMENT_HEADERSIZE;
254	ds1 += LOGFS_SEGMENT_HEADERSIZE;
255	return memcmp(ds0, ds1, len);
256	}
257
258	static int logfs_recover_sb(struct super_block *sb)
259	{
260	struct logfs_super *super = logfs_super(sb);
261	struct logfs_disk_super _ds0, *ds0 = &_ds0;
262	struct logfs_disk_super _ds1, *ds1 = &_ds1;
263	int err, valid0, valid1;
264
265	/* read first superblock */
266	err = wbuf_read(sb, super->s_sb_ofs[0], sizeof(*ds0), ds0);
267	if (err)
268	return err;
269	/* read last superblock */
270	err = wbuf_read(sb, super->s_sb_ofs[1], sizeof(*ds1), ds1);
271	if (err)
272	return err;
273	valid0 = logfs_check_ds(ds0) == 0;
274	valid1 = logfs_check_ds(ds1) == 0;
275
276	if (!valid0 && valid1) {
277	printk(KERN_INFO"First superblock is invalid - fixing.\n");
278	return write_one_sb(sb, super->s_devops->find_first_sb);
279	}
280	if (valid0 && !valid1) {
281	printk(KERN_INFO"Last superblock is invalid - fixing.\n");
282	return write_one_sb(sb, super->s_devops->find_last_sb);
283	}
284	if (valid0 && valid1 && ds_cmp(ds0, ds1)) {
285	printk(KERN_INFO"Superblocks don't match - fixing.\n");
faaa27ab	286	return logfs_write_sb(sb);
5db53f3e JE	287	}
	288	/* If neither is valid now, something's wrong. Didn't we properly
	289	* check them before?!? */
	290	BUG_ON(!valid0 && !valid1);
	291	return 0;
	292	}
	293
	294	static int logfs_make_writeable(struct super_block *sb)
	295	{
	296	int err;
	297
49137f2e JE	298	err = logfs_open_segfile(sb);
	299	if (err)
	300	return err;
	301
5db53f3e JE	302	/* Repair any broken superblock copies */
	303	err = logfs_recover_sb(sb);
	304	if (err)
	305	return err;
	306
	307	/* Check areas for trailing unaccounted data */
	308	err = logfs_check_areas(sb);
	309	if (err)
	310	return err;
	311
5db53f3e JE	312	/* Do one GC pass before any data gets dirtied */
	313	logfs_gc_pass(sb);
	314
	315	/* after all initializations are done, replay the journal
	316	* for rw-mounts, if necessary */
	317	err = logfs_replay_journal(sb);
	318	if (err)
	319	return err;
	320
	321	return 0;
	322	}
	323
	324	static int logfs_get_sb_final(struct super_block sb, struct vfsmount mnt)
	325	{
9421502b	326	struct logfs_super *super = logfs_super(sb);
5db53f3e JE	327	struct inode *rootdir;
	328	int err;
	329
	330	/* root dir */
	331	rootdir = logfs_iget(sb, LOGFS_INO_ROOT);
	332	if (IS_ERR(rootdir))
	333	goto fail;
	334
	335	sb->s_root = d_alloc_root(rootdir);
26562449 AV	336	if (!sb->s_root) {
	337	iput(rootdir);
	338	goto fail;
	339	}
5db53f3e	340
9421502b JE	341	super->s_erase_page = alloc_pages(GFP_KERNEL, 0);
9421502b JE	342	if (!super->s_erase_page)
26562449	343	goto fail;
9421502b JE	344	memset(page_address(super->s_erase_page), 0xFF, PAGE_SIZE);
9421502b JE	345
5db53f3e JE	346	/* FIXME: check for read-only mounts */
	347	err = logfs_make_writeable(sb);
	348	if (err)
26562449	349	goto fail1;
5db53f3e JE	350
	351	log_super("LogFS: Finished mounting\n");
	352	simple_set_mnt(mnt, sb);
	353	return 0;
	354
26562449	355	fail1:
9421502b	356	__free_page(super->s_erase_page);
5db53f3e JE	357	fail:
	358	iput(logfs_super(sb)->s_master_inode);
	359	return -EIO;
	360	}
	361
	362	int logfs_check_ds(struct logfs_disk_super *ds)
	363	{
	364	struct logfs_segment_header *sh = &ds->ds_sh;
	365
	366	if (ds->ds_magic != cpu_to_be64(LOGFS_MAGIC))
	367	return -EINVAL;
	368	if (sh->crc != logfs_crc32(sh, LOGFS_SEGMENT_HEADERSIZE, 4))
	369	return -EINVAL;
	370	if (ds->ds_crc != logfs_crc32(ds, sizeof(*ds),
	371	LOGFS_SEGMENT_HEADERSIZE + 12))
	372	return -EINVAL;
	373	return 0;
	374	}
	375
	376	static struct page find_super_block(struct super_block sb)
	377	{
	378	struct logfs_super *super = logfs_super(sb);
	379	struct page first, last;
	380
	381	first = super->s_devops->find_first_sb(sb, &super->s_sb_ofs[0]);
	382	if (!first \|\| IS_ERR(first))
	383	return NULL;
	384	last = super->s_devops->find_last_sb(sb, &super->s_sb_ofs[1]);
	385	if (!last \|\| IS_ERR(first)) {
	386	page_cache_release(first);
	387	return NULL;
	388	}
	389
	390	if (!logfs_check_ds(page_address(first))) {
	391	page_cache_release(last);
	392	return first;
	393	}
	394
	395	/* First one didn't work, try the second superblock */
	396	if (!logfs_check_ds(page_address(last))) {
	397	page_cache_release(first);
	398	return last;
	399	}
	400
	401	/* Neither worked, sorry folks */
	402	page_cache_release(first);
	403	page_cache_release(last);
	404	return NULL;
	405	}
	406
	407	static int __logfs_read_sb(struct super_block *sb)
	408	{
	409	struct logfs_super *super = logfs_super(sb);
	410	struct page *page;
	411	struct logfs_disk_super *ds;
	412	int i;
	413
	414	page = find_super_block(sb);
	415	if (!page)
	416	return -EIO;
	417
	418	ds = page_address(page);
	419	super->s_size = be64_to_cpu(ds->ds_filesystem_size);
	420	super->s_root_reserve = be64_to_cpu(ds->ds_root_reserve);
421	super->s_speed_reserve = be64_to_cpu(ds->ds_speed_reserve);
422	super->s_bad_seg_reserve = be32_to_cpu(ds->ds_bad_seg_reserve);
423	super->s_segsize = 1 << ds->ds_segment_shift;
424	super->s_segmask = (1 << ds->ds_segment_shift) - 1;
425	super->s_segshift = ds->ds_segment_shift;
426	sb->s_blocksize = 1 << ds->ds_block_shift;
427	sb->s_blocksize_bits = ds->ds_block_shift;
428	super->s_writesize = 1 << ds->ds_write_shift;
429	super->s_writeshift = ds->ds_write_shift;
430	super->s_no_segs = super->s_size >> super->s_segshift;
431	super->s_no_blocks = super->s_segsize >> sb->s_blocksize_bits;
432	super->s_feature_incompat = be64_to_cpu(ds->ds_feature_incompat);
433	super->s_feature_ro_compat = be64_to_cpu(ds->ds_feature_ro_compat);
434	super->s_feature_compat = be64_to_cpu(ds->ds_feature_compat);
435	super->s_feature_flags = be64_to_cpu(ds->ds_feature_flags);
436
437	journal_for_each(i)
438	super->s_journal_seg[i] = be32_to_cpu(ds->ds_journal_seg[i]);
439
440	super->s_ifile_levels = ds->ds_ifile_levels;
441	super->s_iblock_levels = ds->ds_iblock_levels;
442	super->s_data_levels = ds->ds_data_levels;
443	super->s_total_levels = super->s_ifile_levels + super->s_iblock_levels
444	+ super->s_data_levels;
445	page_cache_release(page);
446	return 0;
447	}
448
6a08ab84	449	static int logfs_read_sb(struct super_block *sb, int read_only)
5db53f3e JE	450	{
	451	struct logfs_super *super = logfs_super(sb);
	452	int ret;
	453
	454	super->s_btree_pool = mempool_create(32, btree_alloc, btree_free, NULL);
	455	if (!super->s_btree_pool)
	456	return -ENOMEM;
	457
	458	btree_init_mempool64(&super->s_shadow_tree.new, super->s_btree_pool);
	459	btree_init_mempool64(&super->s_shadow_tree.old, super->s_btree_pool);
032d8f72 JE	460	btree_init_mempool32(&super->s_shadow_tree.segment_map,
032d8f72 JE	461	super->s_btree_pool);
5db53f3e JE	462
	463	ret = logfs_init_mapping(sb);
	464	if (ret)
	465	return ret;
	466
	467	ret = __logfs_read_sb(sb);
	468	if (ret)
	469	return ret;
	470
6a08ab84 JE	471	if (super->s_feature_incompat & ~LOGFS_FEATURES_INCOMPAT)
	472	return -EIO;
	473	if ((super->s_feature_ro_compat & ~LOGFS_FEATURES_RO_COMPAT) &&
	474	!read_only)
	475	return -EIO;
	476
5db53f3e JE	477	mutex_init(&super->s_dirop_mutex);
	478	mutex_init(&super->s_object_alias_mutex);
	479	INIT_LIST_HEAD(&super->s_freeing_list);
	480
	481	ret = logfs_init_rw(sb);
	482	if (ret)
	483	return ret;
	484
	485	ret = logfs_init_areas(sb);
	486	if (ret)
	487	return ret;
	488
	489	ret = logfs_init_gc(sb);
	490	if (ret)
	491	return ret;
	492
	493	ret = logfs_init_journal(sb);
	494	if (ret)
	495	return ret;
	496
	497	return 0;
	498	}
	499
	500	static void logfs_kill_sb(struct super_block *sb)
	501	{
	502	struct logfs_super *super = logfs_super(sb);
	503
	504	log_super("LogFS: Start unmounting\n");
	505	/* Alias entries slow down mount, so evict as many as possible */
	506	sync_filesystem(sb);
c6d38301	507	logfs_write_anchor(sb);
5db53f3e JE	508
	509	/*
	510	* From this point on alias entries are simply dropped - and any
	511	* writes to the object store are considered bugs.
	512	*/
	513	super->s_flags \|= LOGFS_SB_FLAG_SHUTDOWN;
	514	log_super("LogFS: Now in shutdown\n");
	515	generic_shutdown_super(sb);
	516
	517	BUG_ON(super->s_dirty_used_bytes \|\| super->s_dirty_free_bytes);
	518
	519	logfs_cleanup_gc(sb);
	520	logfs_cleanup_journal(sb);
	521	logfs_cleanup_areas(sb);
	522	logfs_cleanup_rw(sb);
9421502b JE	523	if (super->s_erase_page)
9421502b JE	524	__free_page(super->s_erase_page);
5db53f3e	525	super->s_devops->put_device(sb);
1f1b0008 JE	526	logfs_mempool_destroy(super->s_btree_pool);
1f1b0008 JE	527	logfs_mempool_destroy(super->s_alias_pool);
5db53f3e JE	528	kfree(super);
	529	log_super("LogFS: Finished unmounting\n");
	530	}
	531
	532	int logfs_get_sb_device(struct file_system_type *type, int flags,
	533	struct mtd_info mtd, struct block_device bdev,
	534	const struct logfs_device_ops devops, struct vfsmount mnt)
	535	{
	536	struct logfs_super *super;
	537	struct super_block *sb;
	538	int err = -ENOMEM;
	539	static int mount_count;
	540
	541	log_super("LogFS: Start mount %x\n", mount_count++);
	542	super = kzalloc(sizeof(*super), GFP_KERNEL);
	543	if (!super)
	544	goto err0;
	545
	546	super->s_mtd = mtd;
	547	super->s_bdev = bdev;
	548	err = -EINVAL;
	549	sb = sget(type, logfs_sb_test, logfs_sb_set, super);
	550	if (IS_ERR(sb))
	551	goto err0;
	552
	553	if (sb->s_root) {
	554	/* Device is already in use */
	555	err = 0;
	556	simple_set_mnt(mnt, sb);
	557	goto err0;
	558	}
	559
	560	super->s_devops = devops;
	561
	562	/*
	563	* sb->s_maxbytes is limited to 8TB. On 32bit systems, the page cache
	564	* only covers 16TB and the upper 8TB are used for indirect blocks.
	565	* On 64bit system we could bump up the limit, but that would make
	566	* the filesystem incompatible with 32bit systems.
	567	*/
	568	sb->s_maxbytes = (1ull << 43) - 1;
	569	sb->s_op = &logfs_super_operations;
	570	sb->s_flags = flags \| MS_NOATIME;
	571
6a08ab84	572	err = logfs_read_sb(sb, sb->s_flags & MS_RDONLY);
5db53f3e JE	573	if (err)
	574	goto err1;
	575
	576	sb->s_flags \|= MS_ACTIVE;
	577	err = logfs_get_sb_final(sb, mnt);
	578	if (err)
	579	goto err1;
	580	return 0;
	581
	582	err1:
6f2e9e6a	583	deactivate_locked_super(sb);
5db53f3e JE	584	return err;
	585	err0:
	586	kfree(super);
	587	//devops->put_device(sb);
	588	return err;
	589	}
	590
	591	static int logfs_get_sb(struct file_system_type *type, int flags,
	592	const char devname, void data, struct vfsmount *mnt)
	593	{
	594	ulong mtdnr;
	595
	596	if (!devname)
	597	return logfs_get_sb_bdev(type, flags, devname, mnt);
	598	if (strncmp(devname, "mtd", 3))
	599	return logfs_get_sb_bdev(type, flags, devname, mnt);
	600
	601	{
	602	char *garbage;
	603	mtdnr = simple_strtoul(devname+3, &garbage, 0);
	604	if (*garbage)
	605	return -EINVAL;
	606	}
	607
	608	return logfs_get_sb_mtd(type, flags, mtdnr, mnt);
	609	}
	610
	611	static struct file_system_type logfs_fs_type = {
	612	.owner = THIS_MODULE,
	613	.name = "logfs",
	614	.get_sb = logfs_get_sb,
	615	.kill_sb = logfs_kill_sb,
	616	.fs_flags = FS_REQUIRES_DEV,
	617
	618	};
	619
	620	static int __init logfs_init(void)
	621	{
	622	int ret;
	623
	624	emergency_page = alloc_pages(GFP_KERNEL, 0);
	625	if (!emergency_page)
	626	return -ENOMEM;
	627
	628	ret = logfs_compr_init();
	629	if (ret)
	630	goto out1;
	631
	632	ret = logfs_init_inode_cache();
	633	if (ret)
	634	goto out2;
	635
	636	return register_filesystem(&logfs_fs_type);
	637	out2:
	638	logfs_compr_exit();
	639	out1:
	640	__free_pages(emergency_page, 0);
	641	return ret;
	642	}
	643
	644	static void __exit logfs_exit(void)
	645	{
	646	unregister_filesystem(&logfs_fs_type);
	647	logfs_destroy_inode_cache();
648	logfs_compr_exit();
649	__free_pages(emergency_page, 0);
650	}
651
652	module_init(logfs_init);
653	module_exit(logfs_exit);
654
655	MODULE_LICENSE("GPL v2");
656	MODULE_AUTHOR("Joern Engel <joern@logfs.org>");
657	MODULE_DESCRIPTION("scalable flash filesystem");