[linux.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 <[email protected]>
 *
 * 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;
#ifdef CONFIG_BLOCK
	if (sb->s_bdev)
		sb->s_bdi = &bdev_get_queue(sb->s_bdev)->backing_dev_info;
#endif
#ifdef CONFIG_MTD
	if (sb->s_mtd)
		sb->s_bdi = sb->s_mtd->backing_dev_info;
#endif
	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 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;
	}

	/* at that point we know that ->put_super() will be called */
	super->s_erase_page = alloc_pages(GFP_KERNEL, 0);
	if (!super->s_erase_page)
		return -ENOMEM;
	memset(page_address(super->s_erase_page), 0xFF, PAGE_SIZE);

	/* FIXME: check for read-only mounts */
	err = logfs_make_writeable(sb);
	if (err) {
		__free_page(super->s_erase_page);
		return err;
	}

	log_super("LogFS: Finished mounting\n");
	return 0;

fail:
	iput(super->s_master_inode);
	iput(super->s_segfile_inode);
	iput(super->s_mapping_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(last)) {
		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 -EINVAL;

	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;

	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(super);
	logfs_mempool_destroy(super->s_btree_pool);
	logfs_mempool_destroy(super->s_alias_pool);
	kfree(super);
	log_super("LogFS: Finished unmounting\n");
}

static struct dentry *logfs_get_sb_device(struct logfs_super *super,
		struct file_system_type *type, int flags)
{
	struct super_block *sb;
	int err = -ENOMEM;
	static int mount_count;

	log_super("LogFS: Start mount %x\n", mount_count++);

	err = -EINVAL;
	sb = sget(type, logfs_sb_test, logfs_sb_set, super);
	if (IS_ERR(sb)) {
		super->s_devops->put_device(super);
		kfree(super);
		return ERR_CAST(sb);
	}

	if (sb->s_root) {
		/* Device is already in use */
		super->s_devops->put_device(super);
		kfree(super);
		return dget(sb->s_root);
	}

	/*
	 * 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);
	if (err) {
		deactivate_locked_super(sb);
		return ERR_PTR(err);
	}
	return dget(sb->s_root);

err1:
	/* no ->s_root, no ->put_super() */
	iput(super->s_master_inode);
	iput(super->s_segfile_inode);
	iput(super->s_mapping_inode);
	deactivate_locked_super(sb);
	return ERR_PTR(err);
}

static struct dentry *logfs_mount(struct file_system_type *type, int flags,
		const char *devname, void *data)
{
	ulong mtdnr;
	struct logfs_super *super;
	int err;

	super = kzalloc(sizeof(*super), GFP_KERNEL);
	if (!super)
		return ERR_PTR(-ENOMEM);

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

	if (!devname)
		err = logfs_get_sb_bdev(super, type, devname);
	else if (strncmp(devname, "mtd", 3))
		err = logfs_get_sb_bdev(super, type, devname);
	else {
		char *garbage;
		mtdnr = simple_strtoul(devname+3, &garbage, 0);
		if (*garbage)
			err = -EINVAL;
		else
			err = logfs_get_sb_mtd(super, mtdnr);
	}

	if (err) {
		kfree(super);
		return ERR_PTR(err);
	}

	return logfs_get_sb_device(super, type, flags);
}

static struct file_system_type logfs_fs_type = {
	.owner		= THIS_MODULE,
	.name		= "logfs",
	.mount		= logfs_mount,
	.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 <[email protected]>");
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 <[email protected]>
	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;
bba0b5c2	141	#ifdef CONFIG_BLOCK
b8639077 JE	142	if (sb->s_bdev)
b8639077 JE	143	sb->s_bdi = &bdev_get_queue(sb->s_bdev)->backing_dev_info;
bba0b5c2 JE	144	#endif
bba0b5c2 JE	145	#ifdef CONFIG_MTD
b8639077 JE	146	if (sb->s_mtd)
b8639077 JE	147	sb->s_bdi = sb->s_mtd->backing_dev_info;
bba0b5c2	148	#endif
5db53f3e JE	149	return 0;
	150	}
	151
	152	static int logfs_sb_test(struct super_block sb, void _super)
	153	{
	154	struct logfs_super *super = _super;
	155	struct mtd_info *mtd = super->s_mtd;
	156
	157	if (mtd && sb->s_mtd == mtd)
	158	return 1;
	159	if (super->s_bdev && sb->s_bdev == super->s_bdev)
	160	return 1;
	161	return 0;
	162	}
	163
	164	static void set_segment_header(struct logfs_segment_header *sh, u8 type,
	165	u8 level, u32 segno, u32 ec)
	166	{
	167	sh->pad = 0;
	168	sh->type = type;
	169	sh->level = level;
	170	sh->segno = cpu_to_be32(segno);
	171	sh->ec = cpu_to_be32(ec);
	172	sh->gec = cpu_to_be64(segno);
	173	sh->crc = logfs_crc32(sh, LOGFS_SEGMENT_HEADERSIZE, 4);
	174	}
	175
	176	static void logfs_write_ds(struct super_block sb, struct logfs_disk_super ds,
	177	u32 segno, u32 ec)
	178	{
	179	struct logfs_super *super = logfs_super(sb);
	180	struct logfs_segment_header *sh = &ds->ds_sh;
	181	int i;
	182
	183	memset(ds, 0, sizeof(*ds));
	184	set_segment_header(sh, SEG_SUPER, 0, segno, ec);
	185
	186	ds->ds_ifile_levels = super->s_ifile_levels;
	187	ds->ds_iblock_levels = super->s_iblock_levels;
	188	ds->ds_data_levels = super->s_data_levels; /* XXX: Remove */
	189	ds->ds_segment_shift = super->s_segshift;
	190	ds->ds_block_shift = sb->s_blocksize_bits;
	191	ds->ds_write_shift = super->s_writeshift;
	192	ds->ds_filesystem_size = cpu_to_be64(super->s_size);
	193	ds->ds_segment_size = cpu_to_be32(super->s_segsize);
	194	ds->ds_bad_seg_reserve = cpu_to_be32(super->s_bad_seg_reserve);
	195	ds->ds_feature_incompat = cpu_to_be64(super->s_feature_incompat);
	196	ds->ds_feature_ro_compat= cpu_to_be64(super->s_feature_ro_compat);
	197	ds->ds_feature_compat = cpu_to_be64(super->s_feature_compat);
	198	ds->ds_feature_flags = cpu_to_be64(super->s_feature_flags);
	199	ds->ds_root_reserve = cpu_to_be64(super->s_root_reserve);
	200	ds->ds_speed_reserve = cpu_to_be64(super->s_speed_reserve);
	201	journal_for_each(i)
	202	ds->ds_journal_seg[i] = cpu_to_be32(super->s_journal_seg[i]);
	203	ds->ds_magic = cpu_to_be64(LOGFS_MAGIC);
	204	ds->ds_crc = logfs_crc32(ds, sizeof(*ds),
	205	LOGFS_SEGMENT_HEADERSIZE + 12);
	206	}
	207
	208	static int write_one_sb(struct super_block *sb,
	209	struct page (find_sb)(struct super_block sb, u64 ofs))
	210	{
	211	struct logfs_super *super = logfs_super(sb);
	212	struct logfs_disk_super *ds;
213	struct logfs_segment_entry se;
214	struct page *page;
215	u64 ofs;
216	u32 ec, segno;
217	int err;
218
219	page = find_sb(sb, &ofs);
220	if (!page)
221	return -EIO;
222	ds = page_address(page);
223	segno = seg_no(sb, ofs);
224	logfs_get_segment_entry(sb, segno, &se);
225	ec = be32_to_cpu(se.ec_level) >> 4;
226	ec++;
227	logfs_set_segment_erased(sb, segno, ec, 0);
228	logfs_write_ds(sb, ds, segno, ec);
229	err = super->s_devops->write_sb(sb, page);
230	page_cache_release(page);
231	return err;
232	}
233
234	int logfs_write_sb(struct super_block *sb)
235	{
236	struct logfs_super *super = logfs_super(sb);
237	int err;
238
239	/* First superblock */
240	err = write_one_sb(sb, super->s_devops->find_first_sb);
241	if (err)
242	return err;
243
244	/* Last superblock */
245	err = write_one_sb(sb, super->s_devops->find_last_sb);
246	if (err)
247	return err;
248	return 0;
249	}
250
251	static int ds_cmp(const void ds0, const void ds1)
252	{
253	size_t len = sizeof(struct logfs_disk_super);
254
255	/* We know the segment headers differ, so ignore them */
256	len -= LOGFS_SEGMENT_HEADERSIZE;
257	ds0 += LOGFS_SEGMENT_HEADERSIZE;
258	ds1 += LOGFS_SEGMENT_HEADERSIZE;
259	return memcmp(ds0, ds1, len);
260	}
261
262	static int logfs_recover_sb(struct super_block *sb)
263	{
264	struct logfs_super *super = logfs_super(sb);
265	struct logfs_disk_super _ds0, *ds0 = &_ds0;
266	struct logfs_disk_super _ds1, *ds1 = &_ds1;
267	int err, valid0, valid1;
268
269	/* read first superblock */
270	err = wbuf_read(sb, super->s_sb_ofs[0], sizeof(*ds0), ds0);
271	if (err)
272	return err;
273	/* read last superblock */
274	err = wbuf_read(sb, super->s_sb_ofs[1], sizeof(*ds1), ds1);
275	if (err)
276	return err;
277	valid0 = logfs_check_ds(ds0) == 0;
278	valid1 = logfs_check_ds(ds1) == 0;
279
280	if (!valid0 && valid1) {
281	printk(KERN_INFO"First superblock is invalid - fixing.\n");
282	return write_one_sb(sb, super->s_devops->find_first_sb);
283	}
284	if (valid0 && !valid1) {
285	printk(KERN_INFO"Last superblock is invalid - fixing.\n");
286	return write_one_sb(sb, super->s_devops->find_last_sb);
287	}
288	if (valid0 && valid1 && ds_cmp(ds0, ds1)) {
289	printk(KERN_INFO"Superblocks don't match - fixing.\n");
faaa27ab	290	return logfs_write_sb(sb);
5db53f3e JE	291	}
	292	/* If neither is valid now, something's wrong. Didn't we properly
	293	* check them before?!? */
	294	BUG_ON(!valid0 && !valid1);
	295	return 0;
	296	}
	297
	298	static int logfs_make_writeable(struct super_block *sb)
	299	{
	300	int err;
	301
49137f2e JE	302	err = logfs_open_segfile(sb);
	303	if (err)
	304	return err;
	305
5db53f3e JE	306	/* Repair any broken superblock copies */
	307	err = logfs_recover_sb(sb);
	308	if (err)
	309	return err;
	310
	311	/* Check areas for trailing unaccounted data */
	312	err = logfs_check_areas(sb);
	313	if (err)
	314	return err;
	315
5db53f3e JE	316	/* Do one GC pass before any data gets dirtied */
	317	logfs_gc_pass(sb);
	318
	319	/* after all initializations are done, replay the journal
	320	* for rw-mounts, if necessary */
	321	err = logfs_replay_journal(sb);
	322	if (err)
	323	return err;
	324
	325	return 0;
	326	}
	327
a1da9e8a	328	static int logfs_get_sb_final(struct super_block *sb)
5db53f3e	329	{
9421502b	330	struct logfs_super *super = logfs_super(sb);
5db53f3e JE	331	struct inode *rootdir;
	332	int err;
	333
	334	/* root dir */
	335	rootdir = logfs_iget(sb, LOGFS_INO_ROOT);
	336	if (IS_ERR(rootdir))
	337	goto fail;
	338
	339	sb->s_root = d_alloc_root(rootdir);
26562449 AV	340	if (!sb->s_root) {
	341	iput(rootdir);
	342	goto fail;
	343	}
5db53f3e	344
8e22c1a4	345	/* at that point we know that ->put_super() will be called */
9421502b JE	346	super->s_erase_page = alloc_pages(GFP_KERNEL, 0);
9421502b JE	347	if (!super->s_erase_page)
8e22c1a4	348	return -ENOMEM;
9421502b JE	349	memset(page_address(super->s_erase_page), 0xFF, PAGE_SIZE);
9421502b JE	350
5db53f3e JE	351	/* FIXME: check for read-only mounts */
5db53f3e JE	352	err = logfs_make_writeable(sb);
8e22c1a4 AV	353	if (err) {
	354	__free_page(super->s_erase_page);
	355	return err;
	356	}
5db53f3e JE	357
5db53f3e JE	358	log_super("LogFS: Finished mounting\n");
5db53f3e JE	359	return 0;
5db53f3e JE	360
5db53f3e	361	fail:
8e22c1a4 AV	362	iput(super->s_master_inode);
	363	iput(super->s_segfile_inode);
	364	iput(super->s_mapping_inode);
5db53f3e JE	365	return -EIO;
	366	}
	367
	368	int logfs_check_ds(struct logfs_disk_super *ds)
	369	{
	370	struct logfs_segment_header *sh = &ds->ds_sh;
	371
	372	if (ds->ds_magic != cpu_to_be64(LOGFS_MAGIC))
	373	return -EINVAL;
	374	if (sh->crc != logfs_crc32(sh, LOGFS_SEGMENT_HEADERSIZE, 4))
	375	return -EINVAL;
	376	if (ds->ds_crc != logfs_crc32(ds, sizeof(*ds),
	377	LOGFS_SEGMENT_HEADERSIZE + 12))
	378	return -EINVAL;
	379	return 0;
	380	}
	381
	382	static struct page find_super_block(struct super_block sb)
	383	{
	384	struct logfs_super *super = logfs_super(sb);
	385	struct page first, last;
	386
	387	first = super->s_devops->find_first_sb(sb, &super->s_sb_ofs[0]);
	388	if (!first \|\| IS_ERR(first))
	389	return NULL;
	390	last = super->s_devops->find_last_sb(sb, &super->s_sb_ofs[1]);
3272c8a5	391	if (!last \|\| IS_ERR(last)) {
5db53f3e JE	392	page_cache_release(first);
	393	return NULL;
	394	}
	395
	396	if (!logfs_check_ds(page_address(first))) {
	397	page_cache_release(last);
	398	return first;
	399	}
	400
	401	/* First one didn't work, try the second superblock */
	402	if (!logfs_check_ds(page_address(last))) {
	403	page_cache_release(first);
	404	return last;
	405	}
	406
	407	/* Neither worked, sorry folks */
	408	page_cache_release(first);
	409	page_cache_release(last);
	410	return NULL;
	411	}
	412
	413	static int __logfs_read_sb(struct super_block *sb)
	414	{
	415	struct logfs_super *super = logfs_super(sb);
	416	struct page *page;
	417	struct logfs_disk_super *ds;
	418	int i;
	419
	420	page = find_super_block(sb);
	421	if (!page)
ad342631	422	return -EINVAL;
5db53f3e JE	423
	424	ds = page_address(page);
	425	super->s_size = be64_to_cpu(ds->ds_filesystem_size);
	426	super->s_root_reserve = be64_to_cpu(ds->ds_root_reserve);
	427	super->s_speed_reserve = be64_to_cpu(ds->ds_speed_reserve);
	428	super->s_bad_seg_reserve = be32_to_cpu(ds->ds_bad_seg_reserve);
	429	super->s_segsize = 1 << ds->ds_segment_shift;
	430	super->s_segmask = (1 << ds->ds_segment_shift) - 1;
	431	super->s_segshift = ds->ds_segment_shift;
	432	sb->s_blocksize = 1 << ds->ds_block_shift;
	433	sb->s_blocksize_bits = ds->ds_block_shift;
	434	super->s_writesize = 1 << ds->ds_write_shift;
	435	super->s_writeshift = ds->ds_write_shift;
	436	super->s_no_segs = super->s_size >> super->s_segshift;
	437	super->s_no_blocks = super->s_segsize >> sb->s_blocksize_bits;
	438	super->s_feature_incompat = be64_to_cpu(ds->ds_feature_incompat);
	439	super->s_feature_ro_compat = be64_to_cpu(ds->ds_feature_ro_compat);
	440	super->s_feature_compat = be64_to_cpu(ds->ds_feature_compat);
	441	super->s_feature_flags = be64_to_cpu(ds->ds_feature_flags);
	442
	443	journal_for_each(i)
	444	super->s_journal_seg[i] = be32_to_cpu(ds->ds_journal_seg[i]);
	445
	446	super->s_ifile_levels = ds->ds_ifile_levels;
	447	super->s_iblock_levels = ds->ds_iblock_levels;
	448	super->s_data_levels = ds->ds_data_levels;
	449	super->s_total_levels = super->s_ifile_levels + super->s_iblock_levels
	450	+ super->s_data_levels;
	451	page_cache_release(page);
	452	return 0;
	453	}
	454
6a08ab84	455	static int logfs_read_sb(struct super_block *sb, int read_only)
5db53f3e JE	456	{
	457	struct logfs_super *super = logfs_super(sb);
	458	int ret;
	459
	460	super->s_btree_pool = mempool_create(32, btree_alloc, btree_free, NULL);
	461	if (!super->s_btree_pool)
	462	return -ENOMEM;
	463
	464	btree_init_mempool64(&super->s_shadow_tree.new, super->s_btree_pool);
	465	btree_init_mempool64(&super->s_shadow_tree.old, super->s_btree_pool);
032d8f72 JE	466	btree_init_mempool32(&super->s_shadow_tree.segment_map,
032d8f72 JE	467	super->s_btree_pool);
5db53f3e JE	468
	469	ret = logfs_init_mapping(sb);
	470	if (ret)
	471	return ret;
	472
	473	ret = __logfs_read_sb(sb);
	474	if (ret)
	475	return ret;
	476
6a08ab84 JE	477	if (super->s_feature_incompat & ~LOGFS_FEATURES_INCOMPAT)
	478	return -EIO;
	479	if ((super->s_feature_ro_compat & ~LOGFS_FEATURES_RO_COMPAT) &&
	480	!read_only)
	481	return -EIO;
	482
5db53f3e JE	483	ret = logfs_init_rw(sb);
	484	if (ret)
	485	return ret;
	486
	487	ret = logfs_init_areas(sb);
	488	if (ret)
	489	return ret;
	490
	491	ret = logfs_init_gc(sb);
	492	if (ret)
	493	return ret;
	494
	495	ret = logfs_init_journal(sb);
	496	if (ret)
	497	return ret;
	498
	499	return 0;
	500	}
	501
	502	static void logfs_kill_sb(struct super_block *sb)
	503	{
	504	struct logfs_super *super = logfs_super(sb);
	505
	506	log_super("LogFS: Start unmounting\n");
	507	/* Alias entries slow down mount, so evict as many as possible */
	508	sync_filesystem(sb);
c6d38301	509	logfs_write_anchor(sb);
5db53f3e JE	510
	511	/*
	512	* From this point on alias entries are simply dropped - and any
	513	* writes to the object store are considered bugs.
	514	*/
	515	super->s_flags \|= LOGFS_SB_FLAG_SHUTDOWN;
	516	log_super("LogFS: Now in shutdown\n");
	517	generic_shutdown_super(sb);
	518
	519	BUG_ON(super->s_dirty_used_bytes \|\| super->s_dirty_free_bytes);
	520
	521	logfs_cleanup_gc(sb);
	522	logfs_cleanup_journal(sb);
	523	logfs_cleanup_areas(sb);
	524	logfs_cleanup_rw(sb);
9421502b JE	525	if (super->s_erase_page)
9421502b JE	526	__free_page(super->s_erase_page);
e5a0726a	527	super->s_devops->put_device(super);
1f1b0008 JE	528	logfs_mempool_destroy(super->s_btree_pool);
1f1b0008 JE	529	logfs_mempool_destroy(super->s_alias_pool);
5db53f3e JE	530	kfree(super);
	531	log_super("LogFS: Finished unmounting\n");
	532	}
	533
a1da9e8a AV	534	static struct dentry logfs_get_sb_device(struct logfs_super super,
a1da9e8a AV	535	struct file_system_type *type, int flags)
5db53f3e	536	{
5db53f3e JE	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++);
5db53f3e	542
5db53f3e JE	543	err = -EINVAL;
5db53f3e JE	544	sb = sget(type, logfs_sb_test, logfs_sb_set, super);
a1da9e8a AV	545	if (IS_ERR(sb)) {
	546	super->s_devops->put_device(super);
	547	kfree(super);
	548	return ERR_CAST(sb);
	549	}
5db53f3e JE	550
	551	if (sb->s_root) {
	552	/* Device is already in use */
a1da9e8a AV	553	super->s_devops->put_device(super);
	554	kfree(super);
	555	return dget(sb->s_root);
5db53f3e JE	556	}
5db53f3e JE	557
5db53f3e JE	558	/*
	559	* sb->s_maxbytes is limited to 8TB. On 32bit systems, the page cache
	560	* only covers 16TB and the upper 8TB are used for indirect blocks.
	561	* On 64bit system we could bump up the limit, but that would make
	562	* the filesystem incompatible with 32bit systems.
	563	*/
	564	sb->s_maxbytes = (1ull << 43) - 1;
	565	sb->s_op = &logfs_super_operations;
	566	sb->s_flags = flags \| MS_NOATIME;
	567
6a08ab84	568	err = logfs_read_sb(sb, sb->s_flags & MS_RDONLY);
5db53f3e JE	569	if (err)
	570	goto err1;
	571
	572	sb->s_flags \|= MS_ACTIVE;
a1da9e8a AV	573	err = logfs_get_sb_final(sb);
a1da9e8a AV	574	if (err) {
8e22c1a4	575	deactivate_locked_super(sb);
a1da9e8a AV	576	return ERR_PTR(err);
	577	}
	578	return dget(sb->s_root);
5db53f3e JE	579
5db53f3e JE	580	err1:
8e22c1a4 AV	581	/* no ->s_root, no ->put_super() */
	582	iput(super->s_master_inode);
	583	iput(super->s_segfile_inode);
	584	iput(super->s_mapping_inode);
6f2e9e6a	585	deactivate_locked_super(sb);
a1da9e8a	586	return ERR_PTR(err);
5db53f3e JE	587	}
5db53f3e JE	588
a1da9e8a AV	589	static struct dentry logfs_mount(struct file_system_type type, int flags,
a1da9e8a AV	590	const char devname, void data)
5db53f3e JE	591	{
5db53f3e JE	592	ulong mtdnr;
71a1c012	593	struct logfs_super *super;
7d945a3a	594	int err;
71a1c012 AV	595
	596	super = kzalloc(sizeof(*super), GFP_KERNEL);
	597	if (!super)
a1da9e8a	598	return ERR_PTR(-ENOMEM);
5db53f3e	599
cce2c56e LT	600	mutex_init(&super->s_dirop_mutex);
	601	mutex_init(&super->s_object_alias_mutex);
	602	INIT_LIST_HEAD(&super->s_freeing_list);
	603
5db53f3e	604	if (!devname)
7d945a3a AV	605	err = logfs_get_sb_bdev(super, type, devname);
	606	else if (strncmp(devname, "mtd", 3))
	607	err = logfs_get_sb_bdev(super, type, devname);
	608	else {
5db53f3e JE	609	char *garbage;
5db53f3e JE	610	mtdnr = simple_strtoul(devname+3, &garbage, 0);
7d945a3a AV	611	if (*garbage)
	612	err = -EINVAL;
	613	else
	614	err = logfs_get_sb_mtd(super, mtdnr);
	615	}
	616
	617	if (err) {
	618	kfree(super);
a1da9e8a	619	return ERR_PTR(err);
5db53f3e JE	620	}
5db53f3e JE	621
a1da9e8a	622	return logfs_get_sb_device(super, type, flags);
5db53f3e JE	623	}
	624
	625	static struct file_system_type logfs_fs_type = {
	626	.owner = THIS_MODULE,
	627	.name = "logfs",
a1da9e8a	628	.mount = logfs_mount,
5db53f3e JE	629	.kill_sb = logfs_kill_sb,
	630	.fs_flags = FS_REQUIRES_DEV,
	631
	632	};
	633
	634	static int __init logfs_init(void)
	635	{
	636	int ret;
	637
	638	emergency_page = alloc_pages(GFP_KERNEL, 0);
	639	if (!emergency_page)
	640	return -ENOMEM;
	641
	642	ret = logfs_compr_init();
	643	if (ret)
	644	goto out1;
	645
	646	ret = logfs_init_inode_cache();
	647	if (ret)
	648	goto out2;
	649
	650	return register_filesystem(&logfs_fs_type);
	651	out2:
	652	logfs_compr_exit();
	653	out1:
	654	__free_pages(emergency_page, 0);
	655	return ret;
	656	}
	657
	658	static void __exit logfs_exit(void)
	659	{
	660	unregister_filesystem(&logfs_fs_type);
	661	logfs_destroy_inode_cache();
	662	logfs_compr_exit();
	663	__free_pages(emergency_page, 0);
	664	}
	665
	666	module_init(logfs_init);
	667	module_exit(logfs_exit);
	668
	669	MODULE_LICENSE("GPL v2");
	670	MODULE_AUTHOR("Joern Engel <[email protected]>");
	671	MODULE_DESCRIPTION("scalable flash filesystem");