static sector_t find_block(struct inode *inode, struct omfs_extent_entry *ent,
sector_t block, int count, int *left)
{
/* count > 1 because of terminator */
sector_t searched = 0;
for (; count > 1; count--) {
int numblocks = clus_to_blk(OMFS_SB(inode->i_sb),
be64_to_cpu(ent->e_blocks));
if (block >= searched &&
block < searched + numblocks) {
/*
* found it at cluster + (block - searched)
* numblocks - (block - searched) is remainder
*/
*left = numblocks - (block - searched);
return clus_to_blk(OMFS_SB(inode->i_sb),
be64_to_cpu(ent->e_cluster)) +
block - searched;
}
searched += numblocks;
ent++;
}
return 0;
}
/*
* Tries to allocate exactly one block. Returns true if successful.
*/
int omfs_allocate_block(struct super_block *sb, u64 block)
{
struct buffer_head *bh;
struct omfs_sb_info *sbi = OMFS_SB(sb);
int bits_per_entry = 8 * sb->s_blocksize;
unsigned int map, bit;
int ret = 0;
u64 tmp;
tmp = block;
bit = do_div(tmp, bits_per_entry);
map = tmp;
mutex_lock(&sbi->s_bitmap_lock);
if (map >= sbi->s_imap_size || test_and_set_bit(bit, sbi->s_imap[map]))
goto out;
if (sbi->s_bitmap_ino > 0) {
bh = sb_bread(sb, clus_to_blk(sbi, sbi->s_bitmap_ino) + map);
if (!bh)
goto out;
set_bit(bit, (unsigned long *)bh->b_data);
mark_buffer_dirty(bh);
brelse(bh);
}
ret = 1;
out:
mutex_unlock(&sbi->s_bitmap_lock);
return ret;
}
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;
}
struct buffer_head *omfs_bread(struct super_block *sb, sector_t block)
{
struct omfs_sb_info *sbi = OMFS_SB(sb);
if (block >= sbi->s_num_blocks)
return NULL;
return sb_bread(sb, clus_to_blk(sbi, block));
}
/*
* 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);
if (block >= sbi->s_num_blocks)
goto nomem;
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;
}
unsigned long omfs_count_free(struct super_block *sb)
{
unsigned int i;
unsigned long sum = 0;
struct omfs_sb_info *sbi = OMFS_SB(sb);
int nbits = sb->s_blocksize * 8;
for (i = 0; i < sbi->s_imap_size; i++)
sum += nbits - bitmap_weight(sbi->s_imap[i], nbits);
return sum;
}
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);
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_bfree = buf->f_bavail = buf->f_ffree =
omfs_count_free(s);
return 0;
}
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->i_mode = mode;
inode->i_uid = current_fsuid();
inode->i_gid = current_fsgid();
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);
}
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;
}
/*
* Clears count bits starting at a given block.
*/
int omfs_clear_range(struct super_block *sb, u64 block, int count)
{
struct omfs_sb_info *sbi = OMFS_SB(sb);
int bits_per_entry = 8 * sb->s_blocksize;
u64 tmp;
unsigned int map, bit;
int ret;
tmp = block;
bit = do_div(tmp, bits_per_entry);
map = tmp;
if (map >= sbi->s_imap_size)
return 0;
mutex_lock(&sbi->s_bitmap_lock);
ret = set_run(sb, map, bits_per_entry, bit, count, 0);
mutex_unlock(&sbi->s_bitmap_lock);
return ret;
}
/*
* Tries to allocate a set of blocks. The request size depends on the
* type: for inodes, we must allocate sbi->s_mirrors blocks, and for file
* blocks, we try to allocate sbi->s_clustersize, but can always get away
* with just one block.
*/
int omfs_allocate_range(struct super_block *sb,
int min_request,
int max_request,
u64 *return_block,
int *return_size)
{
struct omfs_sb_info *sbi = OMFS_SB(sb);
int bits_per_entry = 8 * sb->s_blocksize;
int ret = 0;
int i, run, bit;
mutex_lock(&sbi->s_bitmap_lock);
for (i = 0; i < sbi->s_imap_size; i++) {
bit = 0;
while (bit < bits_per_entry) {
bit = find_next_zero_bit(sbi->s_imap[i], bits_per_entry,
bit);
if (bit == bits_per_entry)
break;
run = count_run(&sbi->s_imap[i], bits_per_entry,
sbi->s_imap_size-i, bit, max_request);
if (run >= min_request)
goto found;
bit += run;
}
}
ret = -ENOSPC;
goto out;
found:
*return_block = i * bits_per_entry + bit;
*return_size = run;
ret = set_run(sb, i, bits_per_entry, bit, run, 1);
out:
mutex_unlock(&sbi->s_bitmap_lock);
return ret;
}
/*
* Sets or clears the run of count bits starting with bit.
* Called with bitmap lock.
*/
static int set_run(struct super_block *sb, int map,
int nbits, int bit, int count, int set)
{
int i;
int err;
struct buffer_head *bh;
struct omfs_sb_info *sbi = OMFS_SB(sb);
err = -ENOMEM;
bh = sb_bread(sb, clus_to_blk(sbi, sbi->s_bitmap_ino) + map);
if (!bh)
goto out;
for (i = 0; i < count; i++, bit++) {
if (bit >= nbits) {
bit = 0;
map++;
mark_buffer_dirty(bh);
brelse(bh);
bh = sb_bread(sb,
clus_to_blk(sbi, sbi->s_bitmap_ino) + map);
if (!bh)
goto out;
}
if (set) {
set_bit(bit, sbi->s_imap[map]);
set_bit(bit, (unsigned long *)bh->b_data);
} else {
clear_bit(bit, sbi->s_imap[map]);
clear_bit(bit, (unsigned long *)bh->b_data);
}
}
mark_buffer_dirty(bh);
brelse(bh);
err = 0;
out:
return err;
}
/*
* Display the mount options in /proc/mounts.
*/
static int omfs_show_options(struct seq_file *m, struct dentry *root)
{
struct omfs_sb_info *sbi = OMFS_SB(root->d_sb);
umode_t cur_umask = current_umask();
if (!uid_eq(sbi->s_uid, current_uid()))
seq_printf(m, ",uid=%u",
from_kuid_munged(&init_user_ns, sbi->s_uid));
if (!gid_eq(sbi->s_gid, current_gid()))
seq_printf(m, ",gid=%u",
from_kgid_munged(&init_user_ns, sbi->s_gid));
if (sbi->s_dmask == sbi->s_fmask) {
if (sbi->s_fmask != cur_umask)
seq_printf(m, ",umask=%o", sbi->s_fmask);
} else {
if (sbi->s_dmask != cur_umask)
seq_printf(m, ",dmask=%o", sbi->s_dmask);
if (sbi->s_fmask != cur_umask)
seq_printf(m, ",fmask=%o", sbi->s_fmask);
}
return 0;
}
static int omfs_grow_extent(struct inode *inode, struct omfs_extent *oe,
u64 *ret_block)
{
struct omfs_extent_entry *terminator;
struct omfs_extent_entry *entry = &oe->e_entry;
struct omfs_sb_info *sbi = OMFS_SB(inode->i_sb);
u32 extent_count = be32_to_cpu(oe->e_extent_count);
u64 new_block = 0;
u32 max_count;
int new_count;
int ret = 0;
/* reached the end of the extent table with no blocks mapped.
* there are three possibilities for adding: grow last extent,
* add a new extent to the current extent table, and add a
* continuation inode. in last two cases need an allocator for
* sbi->s_cluster_size
*/
/* TODO: handle holes */
/* should always have a terminator */
if (extent_count < 1)
return -EIO;
/* trivially grow current extent, if next block is not taken */
terminator = entry + extent_count - 1;
if (extent_count > 1) {
entry = terminator-1;
new_block = be64_to_cpu(entry->e_cluster) +
be64_to_cpu(entry->e_blocks);
if (omfs_allocate_block(inode->i_sb, new_block)) {
entry->e_blocks =
cpu_to_be64(be64_to_cpu(entry->e_blocks) + 1);
terminator->e_blocks = ~(cpu_to_be64(
be64_to_cpu(~terminator->e_blocks) + 1));
goto out;
}
}
max_count = omfs_max_extents(sbi, OMFS_EXTENT_START);
/* TODO: add a continuation block here */
if (be32_to_cpu(oe->e_extent_count) > max_count-1)
return -EIO;
/* try to allocate a new cluster */
ret = omfs_allocate_range(inode->i_sb, 1, sbi->s_clustersize,
&new_block, &new_count);
if (ret)
goto out_fail;
/* copy terminator down an entry */
entry = terminator;
terminator++;
memcpy(terminator, entry, sizeof(struct omfs_extent_entry));
entry->e_cluster = cpu_to_be64(new_block);
entry->e_blocks = cpu_to_be64((u64) new_count);
terminator->e_blocks = ~(cpu_to_be64(
be64_to_cpu(~terminator->e_blocks) + (u64) new_count));
/* write in new entry */
oe->e_extent_count = cpu_to_be32(1 + be32_to_cpu(oe->e_extent_count));
out:
*ret_block = new_block;
out_fail:
return ret;
}
static int omfs_get_block(struct inode *inode, sector_t block,
struct buffer_head *bh_result, int create)
{
struct buffer_head *bh;
sector_t next, offset;
int ret;
u64 new_block;
u32 max_extents;
int extent_count;
struct omfs_extent *oe;
struct omfs_extent_entry *entry;
struct omfs_sb_info *sbi = OMFS_SB(inode->i_sb);
int max_blocks = bh_result->b_size >> inode->i_blkbits;
int remain;
ret = -EIO;
bh = sb_bread(inode->i_sb, clus_to_blk(sbi, inode->i_ino));
if (!bh)
goto out;
oe = (struct omfs_extent *)(&bh->b_data[OMFS_EXTENT_START]);
max_extents = omfs_max_extents(sbi, OMFS_EXTENT_START);
next = inode->i_ino;
for (;;) {
if (omfs_is_bad(sbi, (struct omfs_header *) bh->b_data, next))
goto out_brelse;
extent_count = be32_to_cpu(oe->e_extent_count);
next = be64_to_cpu(oe->e_next);
entry = &oe->e_entry;
if (extent_count > max_extents)
goto out_brelse;
offset = find_block(inode, entry, block, extent_count, &remain);
if (offset > 0) {
ret = 0;
map_bh(bh_result, inode->i_sb, offset);
if (remain > max_blocks)
remain = max_blocks;
bh_result->b_size = (remain << inode->i_blkbits);
goto out_brelse;
}
if (next == ~0)
break;
brelse(bh);
bh = sb_bread(inode->i_sb, clus_to_blk(sbi, next));
if (!bh)
goto out;
oe = (struct omfs_extent *) (&bh->b_data[OMFS_EXTENT_CONT]);
max_extents = omfs_max_extents(sbi, OMFS_EXTENT_CONT);
}
if (create) {
ret = omfs_grow_extent(inode, oe, &new_block);
if (ret == 0) {
mark_buffer_dirty(bh);
mark_inode_dirty(inode);
map_bh(bh_result, inode->i_sb,
clus_to_blk(sbi, new_block));
}
}
out_brelse:
brelse(bh);
out:
return ret;
}
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;
u64 ctime;
int i;
int ret = -EIO;
int sync_failed = 0;
/* get current inode since we may have written sibling ptrs etc. */
bh = omfs_bread(inode->i_sb, inode->i_ino);
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 = omfs_bread(inode->i_sb, inode->i_ino + i);
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;
}
int omfs_shrink_inode(struct inode *inode)
{
struct omfs_sb_info *sbi = OMFS_SB(inode->i_sb);
struct omfs_extent *oe;
struct omfs_extent_entry *entry;
struct buffer_head *bh;
u64 next, last;
u32 extent_count;
u32 max_extents;
int ret;
/* traverse extent table, freeing each entry that is greater
* than inode->i_size;
*/
next = inode->i_ino;
/* only support truncate -> 0 for now */
ret = -EIO;
if (inode->i_size != 0)
goto out;
bh = sb_bread(inode->i_sb, clus_to_blk(sbi, next));
if (!bh)
goto out;
oe = (struct omfs_extent *)(&bh->b_data[OMFS_EXTENT_START]);
max_extents = omfs_max_extents(sbi, OMFS_EXTENT_START);
for (;;) {
if (omfs_is_bad(sbi, (struct omfs_header *) bh->b_data, next))
goto out_brelse;
extent_count = be32_to_cpu(oe->e_extent_count);
if (extent_count > max_extents)
goto out_brelse;
last = next;
next = be64_to_cpu(oe->e_next);
entry = &oe->e_entry;
/* ignore last entry as it is the terminator */
for (; extent_count > 1; extent_count--) {
u64 start, count;
start = be64_to_cpu(entry->e_cluster);
count = be64_to_cpu(entry->e_blocks);
omfs_clear_range(inode->i_sb, start, (int) count);
entry++;
}
omfs_make_empty_table(bh, (char *) oe - bh->b_data);
mark_buffer_dirty(bh);
brelse(bh);
if (last != inode->i_ino)
omfs_clear_range(inode->i_sb, last, sbi->s_mirrors);
if (next == ~0)
break;
bh = sb_bread(inode->i_sb, clus_to_blk(sbi, next));
if (!bh)
goto out;
oe = (struct omfs_extent *) (&bh->b_data[OMFS_EXTENT_CONT]);
max_extents = omfs_max_extents(sbi, OMFS_EXTENT_CONT);
}
ret = 0;
out:
return ret;
out_brelse:
brelse(bh);
return ret;
}
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;
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;
bh = omfs_bread(inode->i_sb, ino);
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);
}
