/*
* Will look for holes and unwritten extents in the range starting at
* pos for count bytes (inclusive).
*/
static int ocfs2_check_range_for_holes(struct inode *inode, loff_t pos,
size_t count)
{
int ret = 0;
unsigned int extent_flags;
u32 cpos, clusters, extent_len, phys_cpos;
struct super_block *sb = inode->i_sb;
cpos = pos >> OCFS2_SB(sb)->s_clustersize_bits;
clusters = ocfs2_clusters_for_bytes(sb, pos + count) - cpos;
while (clusters) {
ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len,
&extent_flags);
if (ret < 0) {
mlog_errno(ret);
goto out;
}
if (phys_cpos == 0 || (extent_flags & OCFS2_EXT_UNWRITTEN)) {
ret = 1;
break;
}
if (extent_len > clusters)
extent_len = clusters;
clusters -= extent_len;
cpos += extent_len;
}
out:
return ret;
}
static int __ocfs2_move_extents_range(struct buffer_head *di_bh,
struct ocfs2_move_extents_context *context)
{
int ret = 0, flags, do_defrag, skip = 0;
u32 cpos, phys_cpos, move_start, len_to_move, alloc_size;
u32 len_defraged = 0, defrag_thresh = 0, new_phys_cpos = 0;
struct inode *inode = context->inode;
struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
struct ocfs2_move_extents *range = context->range;
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
if ((inode->i_size == 0) || (range->me_len == 0))
return 0;
if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL)
return 0;
context->refcount_loc = le64_to_cpu(di->i_refcount_loc);
ocfs2_init_dinode_extent_tree(&context->et, INODE_CACHE(inode), di_bh);
ocfs2_init_dealloc_ctxt(&context->dealloc);
/*
* TO-DO XXX:
*
* - xattr extents.
*/
do_defrag = context->auto_defrag;
/*
* extents moving happens in unit of clusters, for the sake
* of simplicity, we may ignore two clusters where 'byte_start'
* and 'byte_start + len' were within.
*/
move_start = ocfs2_clusters_for_bytes(osb->sb, range->me_start);
len_to_move = (range->me_start + range->me_len) >>
osb->s_clustersize_bits;
if (len_to_move >= move_start)
len_to_move -= move_start;
else
len_to_move = 0;
if (do_defrag) {
defrag_thresh = range->me_threshold >> osb->s_clustersize_bits;
if (defrag_thresh <= 1)
goto done;
} else
/*
* A tail_to_skip value > 0 indicates that we're being called from
* ocfs2_file_aio_write(). This has the following implications:
*
* - we don't want to update i_size
* - di_bh will be NULL, which is fine because it's only used in the
* case where we want to update i_size.
* - ocfs2_zero_extend() will then only be filling the hole created
* between i_size and the start of the write.
*/
static int ocfs2_extend_file(struct inode *inode,
struct buffer_head *di_bh,
u64 new_i_size,
size_t tail_to_skip)
{
int ret = 0;
u32 clusters_to_add = 0;
BUG_ON(!tail_to_skip && !di_bh);
/* setattr sometimes calls us like this. */
if (new_i_size == 0)
goto out;
if (i_size_read(inode) == new_i_size)
goto out;
BUG_ON(new_i_size < i_size_read(inode));
if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) {
BUG_ON(tail_to_skip != 0);
goto out_update_size;
}
clusters_to_add = ocfs2_clusters_for_bytes(inode->i_sb, new_i_size) -
OCFS2_I(inode)->ip_clusters;
/*
* protect the pages that ocfs2_zero_extend is going to be
* pulling into the page cache.. we do this before the
* metadata extend so that we don't get into the situation
* where we've extended the metadata but can't get the data
* lock to zero.
*/
ret = ocfs2_data_lock(inode, 1);
if (ret < 0) {
mlog_errno(ret);
goto out;
}
if (clusters_to_add) {
ret = ocfs2_extend_allocation(inode, clusters_to_add);
if (ret < 0) {
mlog_errno(ret);
goto out_unlock;
}
}
/*
* Call this even if we don't add any clusters to the tree. We
* still need to zero the area between the old i_size and the
* new i_size.
*/
ret = ocfs2_zero_extend(inode, (u64)new_i_size - tail_to_skip);
if (ret < 0) {
mlog_errno(ret);
goto out_unlock;
}
out_update_size:
if (!tail_to_skip) {
/* We're being called from ocfs2_setattr() which wants
* us to update i_size */
ret = ocfs2_simple_size_update(inode, di_bh, new_i_size);
if (ret < 0)
mlog_errno(ret);
}
out_unlock:
if (!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)))
ocfs2_data_unlock(inode, 1);
out:
return ret;
}
static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
loff_t *ppos,
size_t count,
int appending,
int *direct_io)
{
int ret = 0, meta_level = appending;
struct inode *inode = dentry->d_inode;
u32 clusters;
loff_t newsize, saved_pos;
/*
* We sample i_size under a read level meta lock to see if our write
* is extending the file, if it is we back off and get a write level
* meta lock.
*/
for(;;) {
ret = ocfs2_meta_lock(inode, NULL, meta_level);
if (ret < 0) {
meta_level = -1;
mlog_errno(ret);
goto out;
}
/* Clear suid / sgid if necessary. We do this here
* instead of later in the write path because
* remove_suid() calls ->setattr without any hint that
* we may have already done our cluster locking. Since
* ocfs2_setattr() *must* take cluster locks to
* proceeed, this will lead us to recursively lock the
* inode. There's also the dinode i_size state which
* can be lost via setattr during extending writes (we
* set inode->i_size at the end of a write. */
if (should_remove_suid(dentry)) {
if (meta_level == 0) {
ocfs2_meta_unlock(inode, meta_level);
meta_level = 1;
continue;
}
ret = ocfs2_write_remove_suid(inode);
if (ret < 0) {
mlog_errno(ret);
goto out_unlock;
}
}
/* work on a copy of ppos until we're sure that we won't have
* to recalculate it due to relocking. */
if (appending) {
saved_pos = i_size_read(inode);
mlog(0, "O_APPEND: inode->i_size=%llu\n", saved_pos);
} else {
saved_pos = *ppos;
}
if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) {
loff_t end = saved_pos + count;
/*
* Skip the O_DIRECT checks if we don't need
* them.
*/
if (!direct_io || !(*direct_io))
break;
/*
* Allowing concurrent direct writes means
* i_size changes wouldn't be synchronized, so
* one node could wind up truncating another
* nodes writes.
*/
if (end > i_size_read(inode)) {
*direct_io = 0;
break;
}
/*
* We don't fill holes during direct io, so
* check for them here. If any are found, the
* caller will have to retake some cluster
* locks and initiate the io as buffered.
*/
ret = ocfs2_check_range_for_holes(inode, saved_pos,
count);
if (ret == 1) {
*direct_io = 0;
ret = 0;
} else if (ret < 0)
mlog_errno(ret);
break;
}
/*
* The rest of this loop is concerned with legacy file
* systems which don't support sparse files.
*/
newsize = count + saved_pos;
mlog(0, "pos=%lld newsize=%lld cursize=%lld\n",
(long long) saved_pos, (long long) newsize,
(long long) i_size_read(inode));
/* No need for a higher level metadata lock if we're
* never going past i_size. */
if (newsize <= i_size_read(inode))
break;
if (meta_level == 0) {
ocfs2_meta_unlock(inode, meta_level);
meta_level = 1;
continue;
}
spin_lock(&OCFS2_I(inode)->ip_lock);
clusters = ocfs2_clusters_for_bytes(inode->i_sb, newsize) -
OCFS2_I(inode)->ip_clusters;
spin_unlock(&OCFS2_I(inode)->ip_lock);
mlog(0, "Writing at EOF, may need more allocation: "
"i_size = %lld, newsize = %lld, need %u clusters\n",
(long long) i_size_read(inode), (long long) newsize,
clusters);
/* We only want to continue the rest of this loop if
* our extend will actually require more
* allocation. */
if (!clusters)
break;
ret = ocfs2_extend_file(inode, NULL, newsize, count);
if (ret < 0) {
if (ret != -ENOSPC)
mlog_errno(ret);
goto out_unlock;
}
break;
}
if (appending)
*ppos = saved_pos;
out_unlock:
ocfs2_meta_unlock(inode, meta_level);
out:
return ret;
}
