static char *ocfs2_fast_symlink_getlink(struct inode *inode,
struct buffer_head **bh)
{
int status;
char *link = NULL;
struct ocfs2_dinode *fe;
mlog_entry_void();
status = ocfs2_read_block(OCFS2_SB(inode->i_sb),
OCFS2_I(inode)->ip_blkno,
bh,
OCFS2_BH_CACHED,
inode);
if (status < 0) {
mlog_errno(status);
link = ERR_PTR(status);
goto bail;
}
fe = (struct ocfs2_dinode *) (*bh)->b_data;
link = (char *) fe->id2.i_symlink;
bail:
mlog_exit(status);
return link;
}
static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
handle_t *handle,
u32 bits_wanted,
u32 min_bits,
u16 *bit_off,
unsigned int *num_bits,
u64 gd_blkno,
u16 *bits_left)
{
int ret;
u16 found;
struct buffer_head *group_bh = NULL;
struct ocfs2_group_desc *gd;
struct inode *alloc_inode = ac->ac_inode;
ret = ocfs2_read_block(OCFS2_SB(alloc_inode->i_sb), gd_blkno,
&group_bh, OCFS2_BH_CACHED, alloc_inode);
if (ret < 0) {
mlog_errno(ret);
return ret;
}
gd = (struct ocfs2_group_desc *) group_bh->b_data;
if (!OCFS2_IS_VALID_GROUP_DESC(gd)) {
OCFS2_RO_ON_INVALID_GROUP_DESC(alloc_inode->i_sb, gd);
ret = -EIO;
goto out;
}
ret = ac->ac_group_search(alloc_inode, group_bh, bits_wanted, min_bits,
bit_off, &found);
if (ret < 0) {
if (ret != -ENOSPC)
mlog_errno(ret);
goto out;
}
*num_bits = found;
ret = ocfs2_alloc_dinode_update_counts(alloc_inode, handle, ac->ac_bh,
*num_bits,
le16_to_cpu(gd->bg_chain));
if (ret < 0) {
mlog_errno(ret);
goto out;
}
ret = ocfs2_block_group_set_bits(handle, alloc_inode, gd, group_bh,
*bit_off, *num_bits);
if (ret < 0)
mlog_errno(ret);
*bits_left = le16_to_cpu(gd->bg_free_bits_count);
out:
brelse(group_bh);
return ret;
}
int ocfs2_init_slot_info(struct ocfs2_super *osb)
{
int status, i;
u64 blkno;
struct inode *inode = NULL;
struct buffer_head *bh = NULL;
struct ocfs2_slot_info *si;
si = kzalloc(sizeof(struct ocfs2_slot_info), GFP_KERNEL);
if (!si) {
status = -ENOMEM;
mlog_errno(status);
goto bail;
}
spin_lock_init(&si->si_lock);
si->si_num_slots = osb->max_slots;
si->si_size = OCFS2_MAX_SLOTS;
for(i = 0; i < si->si_num_slots; i++)
si->si_global_node_nums[i] = OCFS2_INVALID_SLOT;
inode = ocfs2_get_system_file_inode(osb, SLOT_MAP_SYSTEM_INODE,
OCFS2_INVALID_SLOT);
if (!inode) {
status = -EINVAL;
mlog_errno(status);
goto bail;
}
status = ocfs2_extent_map_get_blocks(inode, 0ULL, 1, &blkno, NULL);
if (status < 0) {
mlog_errno(status);
goto bail;
}
status = ocfs2_read_block(osb, blkno, &bh, 0, inode);
if (status < 0) {
mlog_errno(status);
goto bail;
}
si->si_inode = inode;
si->si_bh = bh;
osb->slot_info = si;
bail:
if (status < 0 && si)
ocfs2_free_slot_info(si);
return status;
}
static int ocfs2_write_remove_suid(struct inode *inode)
{
int ret;
struct buffer_head *bh = NULL;
struct ocfs2_inode_info *oi = OCFS2_I(inode);
handle_t *handle;
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
struct ocfs2_dinode *di;
mlog_entry("(Inode %llu, mode 0%o)\n",
(unsigned long long)oi->ip_blkno, inode->i_mode);
handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
if (handle == NULL) {
ret = -ENOMEM;
mlog_errno(ret);
goto out;
}
ret = ocfs2_read_block(osb, oi->ip_blkno, &bh, OCFS2_BH_CACHED, inode);
if (ret < 0) {
mlog_errno(ret);
goto out_trans;
}
ret = ocfs2_journal_access(handle, inode, bh,
OCFS2_JOURNAL_ACCESS_WRITE);
if (ret < 0) {
mlog_errno(ret);
goto out_bh;
}
inode->i_mode &= ~S_ISUID;
if ((inode->i_mode & S_ISGID) && (inode->i_mode & S_IXGRP))
inode->i_mode &= ~S_ISGID;
di = (struct ocfs2_dinode *) bh->b_data;
di->i_mode = cpu_to_le16(inode->i_mode);
ret = ocfs2_journal_dirty(handle, bh);
if (ret < 0)
mlog_errno(ret);
out_bh:
brelse(bh);
out_trans:
ocfs2_commit_trans(osb, handle);
out:
mlog_exit(ret);
return ret;
}
/*
* TODO: this should probably be merged into ocfs2_get_block
*
* However, you now need to pay attention to the cont_prepare_write()
* stuff in ocfs2_get_block (that is, ocfs2_get_block pretty much
* expects never to extend).
*/
struct buffer_head *ocfs2_bread(struct inode *inode,
int block, int *err, int reada)
{
struct buffer_head *bh = NULL;
int tmperr;
u64 p_blkno;
int readflags = OCFS2_BH_CACHED;
if (reada)
readflags |= OCFS2_BH_READAHEAD;
if (((u64)block << inode->i_sb->s_blocksize_bits) >=
i_size_read(inode)) {
BUG_ON(!reada);
return NULL;
}
down_read(&OCFS2_I(inode)->ip_alloc_sem);
tmperr = ocfs2_extent_map_get_blocks(inode, block, &p_blkno, NULL,
NULL);
up_read(&OCFS2_I(inode)->ip_alloc_sem);
if (tmperr < 0) {
mlog_errno(tmperr);
goto fail;
}
tmperr = ocfs2_read_block(OCFS2_SB(inode->i_sb), p_blkno, &bh,
readflags, inode);
if (tmperr < 0)
goto fail;
tmperr = 0;
*err = 0;
return bh;
fail:
if (bh) {
brelse(bh);
bh = NULL;
}
*err = -EIO;
return NULL;
}
static int ocfs2_readpage_inline(struct inode *inode, struct page *page)
{
int ret;
struct buffer_head *di_bh = NULL;
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
BUG_ON(!PageLocked(page));
BUG_ON(!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL));
ret = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno, &di_bh,
OCFS2_BH_CACHED, inode);
if (ret) {
mlog_errno(ret);
goto out;
}
ret = ocfs2_read_inline_data(inode, page, di_bh);
out:
unlock_page(page);
brelse(di_bh);
return ret;
}
int ocfs2_read_group_descriptor(struct inode *inode, struct ocfs2_dinode *di,
u64 gd_blkno, struct buffer_head **bh)
{
int rc;
struct buffer_head *tmp = *bh;
rc = ocfs2_read_block(INODE_CACHE(inode), gd_blkno, &tmp,
ocfs2_validate_group_descriptor);
if (rc)
goto out;
rc = ocfs2_validate_gd_parent(inode->i_sb, di, tmp, 0);
if (rc) {
brelse(tmp);
goto out;
}
/* If ocfs2_read_block() got us a new bh, pass it up. */
if (!*bh)
*bh = tmp;
out:
return rc;
}
static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock,
struct buffer_head *bh_result, int create)
{
int err = -EIO;
int status;
struct ocfs2_dinode *fe = NULL;
struct buffer_head *bh = NULL;
struct buffer_head *buffer_cache_bh = NULL;
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
void *kaddr;
mlog_entry("(0x%p, %llu, 0x%p, %d)\n", inode,
(unsigned long long)iblock, bh_result, create);
BUG_ON(ocfs2_inode_is_fast_symlink(inode));
if ((iblock << inode->i_sb->s_blocksize_bits) > PATH_MAX + 1) {
mlog(ML_ERROR, "block offset > PATH_MAX: %llu",
(unsigned long long)iblock);
goto bail;
}
status = ocfs2_read_block(OCFS2_SB(inode->i_sb),
OCFS2_I(inode)->ip_blkno,
&bh, OCFS2_BH_CACHED, inode);
if (status < 0) {
mlog_errno(status);
goto bail;
}
fe = (struct ocfs2_dinode *) bh->b_data;
if (!OCFS2_IS_VALID_DINODE(fe)) {
mlog(ML_ERROR, "Invalid dinode #%llu: signature = %.*s\n",
(unsigned long long)le64_to_cpu(fe->i_blkno), 7,
fe->i_signature);
goto bail;
}
if ((u64)iblock >= ocfs2_clusters_to_blocks(inode->i_sb,
le32_to_cpu(fe->i_clusters))) {
mlog(ML_ERROR, "block offset is outside the allocated size: "
"%llu\n", (unsigned long long)iblock);
goto bail;
}
/* We don't use the page cache to create symlink data, so if
* need be, copy it over from the buffer cache. */
if (!buffer_uptodate(bh_result) && ocfs2_inode_is_new(inode)) {
u64 blkno = le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) +
iblock;
buffer_cache_bh = sb_getblk(osb->sb, blkno);
if (!buffer_cache_bh) {
mlog(ML_ERROR, "couldn't getblock for symlink!\n");
goto bail;
}
/* we haven't locked out transactions, so a commit
* could've happened. Since we've got a reference on
* the bh, even if it commits while we're doing the
* copy, the data is still good. */
if (buffer_jbd(buffer_cache_bh)
&& ocfs2_inode_is_new(inode)) {
kaddr = kmap_atomic(bh_result->b_page, KM_USER0);
if (!kaddr) {
mlog(ML_ERROR, "couldn't kmap!\n");
goto bail;
}
memcpy(kaddr + (bh_result->b_size * iblock),
buffer_cache_bh->b_data,
bh_result->b_size);
kunmap_atomic(kaddr, KM_USER0);
set_buffer_uptodate(bh_result);
}
brelse(buffer_cache_bh);
}
map_bh(bh_result, inode->i_sb,
le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) + iblock);
err = 0;
bail:
if (bh)
brelse(bh);
mlog_exit(err);
return err;
}
static int ocfs2_read_locked_inode(struct inode *inode,
struct ocfs2_find_inode_args *args)
{
struct super_block *sb;
struct ocfs2_super *osb;
struct ocfs2_dinode *fe;
struct buffer_head *bh = NULL;
int status, can_lock;
u32 generation = 0;
mlog_entry("(0x%p, 0x%p)\n", inode, args);
status = -EINVAL;
if (inode == NULL || inode->i_sb == NULL) {
mlog(ML_ERROR, "bad inode\n");
return status;
}
sb = inode->i_sb;
osb = OCFS2_SB(sb);
if (!args) {
mlog(ML_ERROR, "bad inode args\n");
make_bad_inode(inode);
return status;
}
/*
* To improve performance of cold-cache inode stats, we take
* the cluster lock here if possible.
*
* Generally, OCFS2 never trusts the contents of an inode
* unless it's holding a cluster lock, so taking it here isn't
* a correctness issue as much as it is a performance
* improvement.
*
* There are three times when taking the lock is not a good idea:
*
* 1) During startup, before we have initialized the DLM.
*
* 2) If we are reading certain system files which never get
* cluster locks (local alloc, truncate log).
*
* 3) If the process doing the iget() is responsible for
* orphan dir recovery. We're holding the orphan dir lock and
* can get into a deadlock with another process on another
* node in ->delete_inode().
*
* #1 and #2 can be simply solved by never taking the lock
* here for system files (which are the only type we read
* during mount). It's a heavier approach, but our main
* concern is user-accesible files anyway.
*
* #3 works itself out because we'll eventually take the
* cluster lock before trusting anything anyway.
*/
can_lock = !(args->fi_flags & OCFS2_FI_FLAG_SYSFILE)
&& !(args->fi_flags & OCFS2_FI_FLAG_ORPHAN_RECOVERY)
&& !ocfs2_mount_local(osb);
/*
* To maintain backwards compatibility with older versions of
* ocfs2-tools, we still store the generation value for system
* files. The only ones that actually matter to userspace are
* the journals, but it's easier and inexpensive to just flag
* all system files similarly.
*/
if (args->fi_flags & OCFS2_FI_FLAG_SYSFILE)
generation = osb->fs_generation;
ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_meta_lockres,
OCFS2_LOCK_TYPE_META,
generation, inode);
ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_open_lockres,
OCFS2_LOCK_TYPE_OPEN,
0, inode);
if (can_lock) {
status = ocfs2_open_lock(inode);
if (status) {
make_bad_inode(inode);
mlog_errno(status);
return status;
}
status = ocfs2_meta_lock(inode, NULL, 0);
if (status) {
make_bad_inode(inode);
mlog_errno(status);
return status;
}
}
if (args->fi_flags & OCFS2_FI_FLAG_ORPHAN_RECOVERY) {
status = ocfs2_try_open_lock(inode, 0);
if (status) {
make_bad_inode(inode);
return status;
}
}
status = ocfs2_read_block(osb, args->fi_blkno, &bh, 0,
can_lock ? inode : NULL);
if (status < 0) {
mlog_errno(status);
goto bail;
}
status = -EINVAL;
fe = (struct ocfs2_dinode *) bh->b_data;
if (!OCFS2_IS_VALID_DINODE(fe)) {
mlog(0, "Invalid dinode #%llu: signature = %.*s\n",
(unsigned long long)args->fi_blkno, 7,
fe->i_signature);
goto bail;
}
/*
* This is a code bug. Right now the caller needs to
* understand whether it is asking for a system file inode or
* not so the proper lock names can be built.
*/
mlog_bug_on_msg(!!(fe->i_flags & cpu_to_le32(OCFS2_SYSTEM_FL)) !=
!!(args->fi_flags & OCFS2_FI_FLAG_SYSFILE),
"Inode %llu: system file state is ambigous\n",
(unsigned long long)args->fi_blkno);
if (S_ISCHR(le16_to_cpu(fe->i_mode)) ||
S_ISBLK(le16_to_cpu(fe->i_mode)))
inode->i_rdev = huge_decode_dev(le64_to_cpu(fe->id1.dev1.i_rdev));
if (ocfs2_populate_inode(inode, fe, 0) < 0)
goto bail;
BUG_ON(args->fi_blkno != le64_to_cpu(fe->i_blkno));
status = 0;
bail:
if (can_lock)
ocfs2_meta_unlock(inode, 0);
if (status < 0)
make_bad_inode(inode);
if (args && bh)
brelse(bh);
mlog_exit(status);
return status;
}
/*
* Extend the filesystem to the new number of clusters specified. This entry
* point is only used to extend the current filesystem to the end of the last
* existing group.
*/
int ocfs2_group_extend(struct inode * inode, int new_clusters)
{
int ret;
handle_t *handle;
struct buffer_head *main_bm_bh = NULL;
struct buffer_head *group_bh = NULL;
struct inode *main_bm_inode = NULL;
struct ocfs2_dinode *fe = NULL;
struct ocfs2_group_desc *group = NULL;
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
u16 cl_bpc;
u32 first_new_cluster;
u64 lgd_blkno;
mlog_entry_void();
if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
return -EROFS;
if (new_clusters < 0)
return -EINVAL;
else if (new_clusters == 0)
return 0;
main_bm_inode = ocfs2_get_system_file_inode(osb,
GLOBAL_BITMAP_SYSTEM_INODE,
OCFS2_INVALID_SLOT);
if (!main_bm_inode) {
ret = -EINVAL;
mlog_errno(ret);
goto out;
}
mutex_lock(&main_bm_inode->i_mutex);
ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
if (ret < 0) {
mlog_errno(ret);
goto out_mutex;
}
fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
ocfs2_group_bitmap_size(osb->sb) * 8) {
mlog(ML_ERROR, "The disk is too old and small. "
"Force to do offline resize.");
ret = -EINVAL;
goto out_unlock;
}
if (!OCFS2_IS_VALID_DINODE(fe)) {
OCFS2_RO_ON_INVALID_DINODE(main_bm_inode->i_sb, fe);
ret = -EIO;
goto out_unlock;
}
first_new_cluster = le32_to_cpu(fe->i_clusters);
lgd_blkno = ocfs2_which_cluster_group(main_bm_inode,
first_new_cluster - 1);
ret = ocfs2_read_block(main_bm_inode, lgd_blkno, &group_bh);
if (ret < 0) {
mlog_errno(ret);
goto out_unlock;
}
group = (struct ocfs2_group_desc *)group_bh->b_data;
ret = ocfs2_check_group_descriptor(inode->i_sb, fe, group);
if (ret) {
mlog_errno(ret);
goto out_unlock;
}
cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
if (le16_to_cpu(group->bg_bits) / cl_bpc + new_clusters >
le16_to_cpu(fe->id2.i_chain.cl_cpg)) {
ret = -EINVAL;
goto out_unlock;
}
mlog(0, "extend the last group at %llu, new clusters = %d\n",
(unsigned long long)le64_to_cpu(group->bg_blkno), new_clusters);
handle = ocfs2_start_trans(osb, OCFS2_GROUP_EXTEND_CREDITS);
if (IS_ERR(handle)) {
mlog_errno(PTR_ERR(handle));
ret = -EINVAL;
goto out_unlock;
}
/* update the last group descriptor and inode. */
ret = ocfs2_update_last_group_and_inode(handle, main_bm_inode,
main_bm_bh, group_bh,
first_new_cluster,
new_clusters);
if (ret) {
mlog_errno(ret);
goto out_commit;
}
ocfs2_update_super_and_backups(main_bm_inode, new_clusters);
out_commit:
ocfs2_commit_trans(osb, handle);
out_unlock:
brelse(group_bh);
brelse(main_bm_bh);
ocfs2_inode_unlock(main_bm_inode, 1);
out_mutex:
mutex_unlock(&main_bm_inode->i_mutex);
iput(main_bm_inode);
out:
mlog_exit_void();
return ret;
}
/*
* We want to free the bitmap bits outside of any recovery context as
* we'll need a cluster lock to do so, but we must clear the local
* alloc before giving up the recovered nodes journal. To solve this,
* we kmalloc a copy of the local alloc before it's change for the
* caller to process with ocfs2_complete_local_alloc_recovery
*/
int ocfs2_begin_local_alloc_recovery(struct ocfs2_super *osb,
int slot_num,
struct ocfs2_dinode **alloc_copy)
{
int status = 0;
struct buffer_head *alloc_bh = NULL;
struct inode *inode = NULL;
struct ocfs2_dinode *alloc;
mlog_entry("(slot_num = %d)\n", slot_num);
*alloc_copy = NULL;
inode = ocfs2_get_system_file_inode(osb,
LOCAL_ALLOC_SYSTEM_INODE,
slot_num);
if (!inode) {
status = -EINVAL;
mlog_errno(status);
goto bail;
}
mutex_lock(&inode->i_mutex);
status = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno,
&alloc_bh, 0, inode);
if (status < 0) {
mlog_errno(status);
goto bail;
}
*alloc_copy = kmalloc(alloc_bh->b_size, GFP_KERNEL);
if (!(*alloc_copy)) {
status = -ENOMEM;
goto bail;
}
memcpy((*alloc_copy), alloc_bh->b_data, alloc_bh->b_size);
alloc = (struct ocfs2_dinode *) alloc_bh->b_data;
ocfs2_clear_local_alloc(alloc);
status = ocfs2_write_block(osb, alloc_bh, inode);
if (status < 0)
mlog_errno(status);
bail:
if ((status < 0) && (*alloc_copy)) {
kfree(*alloc_copy);
*alloc_copy = NULL;
}
if (alloc_bh)
brelse(alloc_bh);
if (inode) {
mutex_unlock(&inode->i_mutex);
iput(inode);
}
mlog_exit(status);
return status;
}
static int ocfs2_extend_allocation(struct inode *inode,
u32 clusters_to_add)
{
int status = 0;
int restart_func = 0;
int drop_alloc_sem = 0;
int credits;
u32 prev_clusters, logical_start;
struct buffer_head *bh = NULL;
struct ocfs2_dinode *fe = NULL;
handle_t *handle = NULL;
struct ocfs2_alloc_context *data_ac = NULL;
struct ocfs2_alloc_context *meta_ac = NULL;
enum ocfs2_alloc_restarted why;
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
mlog_entry("(clusters_to_add = %u)\n", clusters_to_add);
/*
* This function only exists for file systems which don't
* support holes.
*/
BUG_ON(ocfs2_sparse_alloc(osb));
status = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno, &bh,
OCFS2_BH_CACHED, inode);
if (status < 0) {
mlog_errno(status);
goto leave;
}
fe = (struct ocfs2_dinode *) bh->b_data;
if (!OCFS2_IS_VALID_DINODE(fe)) {
OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
status = -EIO;
goto leave;
}
logical_start = OCFS2_I(inode)->ip_clusters;
restart_all:
BUG_ON(le32_to_cpu(fe->i_clusters) != OCFS2_I(inode)->ip_clusters);
/* blocks peope in read/write from reading our allocation
* until we're done changing it. We depend on i_mutex to block
* other extend/truncate calls while we're here. Ordering wrt
* start_trans is important here -- always do it before! */
down_write(&OCFS2_I(inode)->ip_alloc_sem);
drop_alloc_sem = 1;
status = ocfs2_lock_allocators(inode, fe, clusters_to_add, &data_ac,
&meta_ac);
if (status) {
mlog_errno(status);
goto leave;
}
credits = ocfs2_calc_extend_credits(osb->sb, fe, clusters_to_add);
handle = ocfs2_start_trans(osb, credits);
if (IS_ERR(handle)) {
status = PTR_ERR(handle);
handle = NULL;
mlog_errno(status);
goto leave;
}
restarted_transaction:
/* reserve a write to the file entry early on - that we if we
* run out of credits in the allocation path, we can still
* update i_size. */
status = ocfs2_journal_access(handle, inode, bh,
OCFS2_JOURNAL_ACCESS_WRITE);
if (status < 0) {
mlog_errno(status);
goto leave;
}
prev_clusters = OCFS2_I(inode)->ip_clusters;
status = ocfs2_do_extend_allocation(osb,
inode,
&logical_start,
clusters_to_add,
bh,
handle,
data_ac,
meta_ac,
&why);
if ((status < 0) && (status != -EAGAIN)) {
if (status != -ENOSPC)
mlog_errno(status);
goto leave;
}
status = ocfs2_journal_dirty(handle, bh);
if (status < 0) {
mlog_errno(status);
goto leave;
}
spin_lock(&OCFS2_I(inode)->ip_lock);
clusters_to_add -= (OCFS2_I(inode)->ip_clusters - prev_clusters);
spin_unlock(&OCFS2_I(inode)->ip_lock);
if (why != RESTART_NONE && clusters_to_add) {
if (why == RESTART_META) {
mlog(0, "restarting function.\n");
restart_func = 1;
} else {
BUG_ON(why != RESTART_TRANS);
mlog(0, "restarting transaction.\n");
/* TODO: This can be more intelligent. */
credits = ocfs2_calc_extend_credits(osb->sb,
fe,
clusters_to_add);
status = ocfs2_extend_trans(handle, credits);
if (status < 0) {
/* handle still has to be committed at
* this point. */
status = -ENOMEM;
mlog_errno(status);
goto leave;
}
goto restarted_transaction;
}
}
mlog(0, "fe: i_clusters = %u, i_size=%llu\n",
le32_to_cpu(fe->i_clusters),
(unsigned long long)le64_to_cpu(fe->i_size));
mlog(0, "inode: ip_clusters=%u, i_size=%lld\n",
OCFS2_I(inode)->ip_clusters, i_size_read(inode));
leave:
if (drop_alloc_sem) {
up_write(&OCFS2_I(inode)->ip_alloc_sem);
drop_alloc_sem = 0;
}
if (handle) {
ocfs2_commit_trans(osb, handle);
handle = NULL;
}
if (data_ac) {
ocfs2_free_alloc_context(data_ac);
data_ac = NULL;
}
if (meta_ac) {
ocfs2_free_alloc_context(meta_ac);
meta_ac = NULL;
}
if ((!status) && restart_func) {
restart_func = 0;
goto restart_all;
}
if (bh) {
brelse(bh);
bh = NULL;
}
mlog_exit(status);
return status;
}
/*
* expects the suballoc inode to already be locked.
*/
static int ocfs2_free_suballoc_bits(handle_t *handle,
struct inode *alloc_inode,
struct buffer_head *alloc_bh,
unsigned int start_bit,
u64 bg_blkno,
unsigned int count)
{
int status = 0;
u32 tmp_used;
struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
struct ocfs2_dinode *fe = (struct ocfs2_dinode *) alloc_bh->b_data;
struct ocfs2_chain_list *cl = &fe->id2.i_chain;
struct buffer_head *group_bh = NULL;
struct ocfs2_group_desc *group;
mlog_entry_void();
if (!OCFS2_IS_VALID_DINODE(fe)) {
OCFS2_RO_ON_INVALID_DINODE(alloc_inode->i_sb, fe);
status = -EIO;
goto bail;
}
BUG_ON((count + start_bit) > ocfs2_bits_per_group(cl));
mlog(0, "%llu: freeing %u bits from group %llu, starting at %u\n",
(unsigned long long)OCFS2_I(alloc_inode)->ip_blkno, count,
(unsigned long long)bg_blkno, start_bit);
status = ocfs2_read_block(osb, bg_blkno, &group_bh, OCFS2_BH_CACHED,
alloc_inode);
if (status < 0) {
mlog_errno(status);
goto bail;
}
group = (struct ocfs2_group_desc *) group_bh->b_data;
status = ocfs2_check_group_descriptor(alloc_inode->i_sb, fe, group);
if (status) {
mlog_errno(status);
goto bail;
}
BUG_ON((count + start_bit) > le16_to_cpu(group->bg_bits));
status = ocfs2_block_group_clear_bits(handle, alloc_inode,
group, group_bh,
start_bit, count);
if (status < 0) {
mlog_errno(status);
goto bail;
}
status = ocfs2_journal_access(handle, alloc_inode, alloc_bh,
OCFS2_JOURNAL_ACCESS_WRITE);
if (status < 0) {
mlog_errno(status);
goto bail;
}
le32_add_cpu(&cl->cl_recs[le16_to_cpu(group->bg_chain)].c_free,
count);
tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used);
fe->id1.bitmap1.i_used = cpu_to_le32(tmp_used - count);
status = ocfs2_journal_dirty(handle, alloc_bh);
if (status < 0) {
mlog_errno(status);
goto bail;
}
bail:
if (group_bh)
brelse(group_bh);
mlog_exit(status);
return status;
}
static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
handle_t *handle,
u32 bits_wanted,
u32 min_bits,
u16 *bit_off,
unsigned int *num_bits,
u64 *bg_blkno,
u16 *bits_left)
{
int status;
u16 chain, tmp_bits;
u32 tmp_used;
u64 next_group;
struct inode *alloc_inode = ac->ac_inode;
struct buffer_head *group_bh = NULL;
struct buffer_head *prev_group_bh = NULL;
struct ocfs2_dinode *fe = (struct ocfs2_dinode *) ac->ac_bh->b_data;
struct ocfs2_chain_list *cl = (struct ocfs2_chain_list *) &fe->id2.i_chain;
struct ocfs2_group_desc *bg;
chain = ac->ac_chain;
mlog(0, "trying to alloc %u bits from chain %u, inode %llu\n",
bits_wanted, chain,
(unsigned long long)OCFS2_I(alloc_inode)->ip_blkno);
status = ocfs2_read_block(OCFS2_SB(alloc_inode->i_sb),
le64_to_cpu(cl->cl_recs[chain].c_blkno),
&group_bh, OCFS2_BH_CACHED, alloc_inode);
if (status < 0) {
mlog_errno(status);
goto bail;
}
bg = (struct ocfs2_group_desc *) group_bh->b_data;
status = ocfs2_check_group_descriptor(alloc_inode->i_sb, fe, bg);
if (status) {
mlog_errno(status);
goto bail;
}
status = -ENOSPC;
/* for now, the chain search is a bit simplistic. We just use
* the 1st group with any empty bits. */
while ((status = ac->ac_group_search(alloc_inode, group_bh,
bits_wanted, min_bits, bit_off,
&tmp_bits)) == -ENOSPC) {
if (!bg->bg_next_group)
break;
if (prev_group_bh) {
brelse(prev_group_bh);
prev_group_bh = NULL;
}
next_group = le64_to_cpu(bg->bg_next_group);
prev_group_bh = group_bh;
group_bh = NULL;
status = ocfs2_read_block(OCFS2_SB(alloc_inode->i_sb),
next_group, &group_bh,
OCFS2_BH_CACHED, alloc_inode);
if (status < 0) {
mlog_errno(status);
goto bail;
}
bg = (struct ocfs2_group_desc *) group_bh->b_data;
status = ocfs2_check_group_descriptor(alloc_inode->i_sb, fe, bg);
if (status) {
mlog_errno(status);
goto bail;
}
}
if (status < 0) {
if (status != -ENOSPC)
mlog_errno(status);
goto bail;
}
mlog(0, "alloc succeeds: we give %u bits from block group %llu\n",
tmp_bits, (unsigned long long)le64_to_cpu(bg->bg_blkno));
*num_bits = tmp_bits;
BUG_ON(*num_bits == 0);
/*
* Keep track of previous block descriptor read. When
* we find a target, if we have read more than X
* number of descriptors, and the target is reasonably
* empty, relink him to top of his chain.
*
* We've read 0 extra blocks and only send one more to
* the transaction, yet the next guy to search has a
* much easier time.
*
* Do this *after* figuring out how many bits we're taking out
* of our target group.
*/
if (ac->ac_allow_chain_relink &&
(prev_group_bh) &&
(ocfs2_block_group_reasonably_empty(bg, *num_bits))) {
status = ocfs2_relink_block_group(handle, alloc_inode,
ac->ac_bh, group_bh,
prev_group_bh, chain);
if (status < 0) {
mlog_errno(status);
goto bail;
}
}
/* Ok, claim our bits now: set the info on dinode, chainlist
* and then the group */
status = ocfs2_journal_access(handle,
alloc_inode,
ac->ac_bh,
OCFS2_JOURNAL_ACCESS_WRITE);
if (status < 0) {
mlog_errno(status);
goto bail;
}
tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used);
fe->id1.bitmap1.i_used = cpu_to_le32(*num_bits + tmp_used);
le32_add_cpu(&cl->cl_recs[chain].c_free, -(*num_bits));
status = ocfs2_journal_dirty(handle,
ac->ac_bh);
if (status < 0) {
mlog_errno(status);
goto bail;
}
status = ocfs2_block_group_set_bits(handle,
alloc_inode,
bg,
group_bh,
*bit_off,
*num_bits);
if (status < 0) {
mlog_errno(status);
goto bail;
}
mlog(0, "Allocated %u bits from suballocator %llu\n", *num_bits,
(unsigned long long)le64_to_cpu(fe->i_blkno));
*bg_blkno = le64_to_cpu(bg->bg_blkno);
*bits_left = le16_to_cpu(bg->bg_free_bits_count);
bail:
if (group_bh)
brelse(group_bh);
if (prev_group_bh)
brelse(prev_group_bh);
mlog_exit(status);
return status;
}
/*
* Find the leaf containing the interval we want. While we're on our
* way down the tree, fill in every record we see at any depth, because
* we might want it later.
*
* Note that this code is run without ip_lock. That's because it
* sleeps while reading. If someone is also filling the extent list at
* the same time we are, we might have to restart.
*/
static int ocfs2_extent_map_find_leaf(struct inode *inode,
u32 cpos, u32 clusters,
struct ocfs2_extent_list *el)
{
int i, ret;
struct buffer_head *eb_bh = NULL;
u64 blkno;
u32 rec_end;
struct ocfs2_extent_block *eb;
struct ocfs2_extent_rec *rec;
/*
* The bh data containing the el cannot change here, because
* we hold alloc_sem. So we can do this without other
* locks.
*/
while (el->l_tree_depth)
{
blkno = 0;
for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
rec = &el->l_recs[i];
rec_end = (le32_to_cpu(rec->e_cpos) +
le32_to_cpu(rec->e_clusters));
ret = -EBADR;
if (rec_end > OCFS2_I(inode)->ip_clusters) {
mlog_errno(ret);
ocfs2_error(inode->i_sb,
"Extent %d at e_blkno %llu of inode %llu goes past ip_clusters of %u\n",
i,
(unsigned long long)le64_to_cpu(rec->e_blkno),
(unsigned long long)OCFS2_I(inode)->ip_blkno,
OCFS2_I(inode)->ip_clusters);
goto out_free;
}
if (rec_end <= cpos) {
ret = ocfs2_extent_map_insert(inode, rec,
le16_to_cpu(el->l_tree_depth));
if (ret && (ret != -EEXIST)) {
mlog_errno(ret);
goto out_free;
}
continue;
}
if ((cpos + clusters) <= le32_to_cpu(rec->e_cpos)) {
ret = ocfs2_extent_map_insert(inode, rec,
le16_to_cpu(el->l_tree_depth));
if (ret && (ret != -EEXIST)) {
mlog_errno(ret);
goto out_free;
}
continue;
}
/*
* We've found a record that matches our
* interval. We don't insert it because we're
* about to traverse it.
*/
/* Check to see if we're stradling */
ret = -ESRCH;
if (!ocfs2_extent_rec_contains_clusters(rec,
cpos,
clusters)) {
mlog_errno(ret);
goto out_free;
}
/*
* If we've already found a record, the el has
* two records covering the same interval.
* EEEK!
*/
ret = -EBADR;
if (blkno) {
mlog_errno(ret);
ocfs2_error(inode->i_sb,
"Multiple extents for (cpos = %u, clusters = %u) on inode %llu; e_blkno %llu and rec %d at e_blkno %llu\n",
cpos, clusters,
(unsigned long long)OCFS2_I(inode)->ip_blkno,
(unsigned long long)blkno, i,
(unsigned long long)le64_to_cpu(rec->e_blkno));
goto out_free;
}
blkno = le64_to_cpu(rec->e_blkno);
}
/*
* We don't support holes, and we're still up
* in the branches, so we'd better have found someone
*/
ret = -EBADR;
if (!blkno) {
ocfs2_error(inode->i_sb,
"No record found for (cpos = %u, clusters = %u) on inode %llu\n",
cpos, clusters,
(unsigned long long)OCFS2_I(inode)->ip_blkno);
mlog_errno(ret);
goto out_free;
}
if (eb_bh) {
brelse(eb_bh);
eb_bh = NULL;
}
ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
blkno, &eb_bh, OCFS2_BH_CACHED,
inode);
if (ret) {
mlog_errno(ret);
goto out_free;
}
eb = (struct ocfs2_extent_block *)eb_bh->b_data;
if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
ret = -EIO;
goto out_free;
}
el = &eb->h_list;
}
BUG_ON(el->l_tree_depth);
for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
rec = &el->l_recs[i];
if ((le32_to_cpu(rec->e_cpos) + le32_to_cpu(rec->e_clusters)) >
OCFS2_I(inode)->ip_clusters) {
ret = -EBADR;
mlog_errno(ret);
ocfs2_error(inode->i_sb,
"Extent %d at e_blkno %llu of inode %llu goes past ip_clusters of %u\n",
i,
(unsigned long long)le64_to_cpu(rec->e_blkno),
(unsigned long long)OCFS2_I(inode)->ip_blkno,
OCFS2_I(inode)->ip_clusters);
return ret;
}
ret = ocfs2_extent_map_insert(inode, rec,
le16_to_cpu(el->l_tree_depth));
if (ret && (ret != -EEXIST)) {
mlog_errno(ret);
goto out_free;
}
}
ret = 0;
out_free:
if (eb_bh)
brelse(eb_bh);
return ret;
}
/*
* This lookup actually will read from disk. It has one invariant:
* It will never re-traverse blocks. This means that all inserts should
* be new regions or more granular regions (both allowed by insert).
*/
static int ocfs2_extent_map_lookup_read(struct inode *inode,
u32 cpos,
u32 clusters,
struct ocfs2_extent_map_entry **ret_ent)
{
int ret;
u64 blkno;
struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
struct ocfs2_extent_map_entry *ent;
struct buffer_head *bh = NULL;
struct ocfs2_extent_block *eb;
struct ocfs2_dinode *di;
struct ocfs2_extent_list *el;
spin_lock(&OCFS2_I(inode)->ip_lock);
ent = ocfs2_extent_map_lookup(em, cpos, clusters, NULL, NULL);
if (ent) {
if (!ent->e_tree_depth) {
spin_unlock(&OCFS2_I(inode)->ip_lock);
*ret_ent = ent;
return 0;
}
blkno = le64_to_cpu(ent->e_rec.e_blkno);
spin_unlock(&OCFS2_I(inode)->ip_lock);
ret = ocfs2_read_block(OCFS2_SB(inode->i_sb), blkno, &bh,
OCFS2_BH_CACHED, inode);
if (ret) {
mlog_errno(ret);
if (bh)
brelse(bh);
return ret;
}
eb = (struct ocfs2_extent_block *)bh->b_data;
if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
brelse(bh);
return -EIO;
}
el = &eb->h_list;
} else {
spin_unlock(&OCFS2_I(inode)->ip_lock);
ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
OCFS2_I(inode)->ip_blkno, &bh,
OCFS2_BH_CACHED, inode);
if (ret) {
mlog_errno(ret);
if (bh)
brelse(bh);
return ret;
}
di = (struct ocfs2_dinode *)bh->b_data;
if (!OCFS2_IS_VALID_DINODE(di)) {
brelse(bh);
OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, di);
return -EIO;
}
el = &di->id2.i_list;
}
ret = ocfs2_extent_map_find_leaf(inode, cpos, clusters, el);
brelse(bh);
if (ret) {
mlog_errno(ret);
return ret;
}
ent = ocfs2_extent_map_lookup(em, cpos, clusters, NULL, NULL);
if (!ent) {
ret = -ESRCH;
mlog_errno(ret);
return ret;
}
/* FIXME: Make sure this isn't a corruption */
BUG_ON(ent->e_tree_depth);
*ret_ent = ent;
return 0;
}
int ocfs2_load_local_alloc(struct ocfs2_super *osb)
{
int status = 0;
struct ocfs2_dinode *alloc = NULL;
struct buffer_head *alloc_bh = NULL;
u32 num_used;
struct inode *inode = NULL;
struct ocfs2_local_alloc *la;
mlog_entry_void();
/* read the alloc off disk */
inode = ocfs2_get_system_file_inode(osb, LOCAL_ALLOC_SYSTEM_INODE,
osb->slot_num);
if (!inode) {
status = -EINVAL;
mlog_errno(status);
goto bail;
}
status = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno,
&alloc_bh, 0, inode);
if (status < 0) {
mlog_errno(status);
goto bail;
}
alloc = (struct ocfs2_dinode *) alloc_bh->b_data;
la = OCFS2_LOCAL_ALLOC(alloc);
if (!(le32_to_cpu(alloc->i_flags) &
(OCFS2_LOCAL_ALLOC_FL|OCFS2_BITMAP_FL))) {
mlog(ML_ERROR, "Invalid local alloc inode, %"MLFu64"\n",
OCFS2_I(inode)->ip_blkno);
status = -EINVAL;
goto bail;
}
if ((la->la_size == 0) ||
(le16_to_cpu(la->la_size) > ocfs2_local_alloc_size(inode->i_sb))) {
mlog(ML_ERROR, "Local alloc size is invalid (la_size = %u)\n",
le16_to_cpu(la->la_size));
status = -EINVAL;
goto bail;
}
/* do a little verification. */
num_used = ocfs2_local_alloc_count_bits(alloc);
/* hopefully the local alloc has always been recovered before
* we load it. */
if (num_used
|| alloc->id1.bitmap1.i_used
|| alloc->id1.bitmap1.i_total
|| la->la_bm_off)
mlog(ML_ERROR, "Local alloc hasn't been recovered!\n"
"found = %u, set = %u, taken = %u, off = %u\n",
num_used, le32_to_cpu(alloc->id1.bitmap1.i_used),
le32_to_cpu(alloc->id1.bitmap1.i_total),
OCFS2_LOCAL_ALLOC(alloc)->la_bm_off);
osb->local_alloc_bh = alloc_bh;
osb->local_alloc_state = OCFS2_LA_ENABLED;
bail:
if (status < 0)
if (alloc_bh)
brelse(alloc_bh);
if (inode)
iput(inode);
mlog_exit(status);
return status;
}
