Btrfs: add hole punching

This patch adds hole punching via fallocate.  Thanks,

Signed-off-by: Josef Bacik <jbacik@fusionio.com>

fs/btrfs/ctree.h

@@ -3250,6 +3250,8 @@ int btrfs_unlink_subvol(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root,
 			struct inode *dir, u64 objectid,
 			const char *name, int name_len);
+int btrfs_truncate_page(struct inode *inode, loff_t from, loff_t len,
+			int front);
 int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root,
 			       struct inode *inode, u64 new_size,
@@ -3323,7 +3325,7 @@ extern const struct file_operations btrfs_file_operations;
 int __btrfs_drop_extents(struct btrfs_trans_handle *trans,
 			 struct btrfs_root *root, struct inode *inode,
 			 struct btrfs_path *path, u64 start, u64 end,
-			 int drop_cache);
+			 u64 *drop_end, int drop_cache);
 int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, struct inode *inode, u64 start,
 		       u64 end, int drop_cache);

fs/btrfs/extent-tree.c

@@ -4132,6 +4132,8 @@ struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_root *root)
 void btrfs_free_block_rsv(struct btrfs_root *root,
 			  struct btrfs_block_rsv *rsv)
 {
+	if (!rsv)
+		return;
 	btrfs_block_rsv_release(root, rsv, (u64)-1);
 	kfree(rsv);
 }

fs/btrfs/file.c

@@ -39,6 +39,7 @@
 #include "tree-log.h"
 #include "locking.h"
 #include "compat.h"
+#include "volumes.h"
 
 /*
  * when auto defrag is enabled we
@@ -584,7 +585,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
 int __btrfs_drop_extents(struct btrfs_trans_handle *trans,
 			 struct btrfs_root *root, struct inode *inode,
 			 struct btrfs_path *path, u64 start, u64 end,
-			 int drop_cache)
+			 u64 *drop_end, int drop_cache)
 {
 	struct extent_buffer *leaf;
 	struct btrfs_file_extent_item *fi;
@@ -822,6 +823,8 @@ int __btrfs_drop_extents(struct btrfs_trans_handle *trans,
 			btrfs_abort_transaction(trans, root, ret);
 	}
 
+	if (drop_end)
+		*drop_end = min(end, extent_end);
 	btrfs_release_path(path);
 	return ret;
 }
@@ -836,7 +839,7 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
-	ret = __btrfs_drop_extents(trans, root, inode, path, start, end,
+	ret = __btrfs_drop_extents(trans, root, inode, path, start, end, NULL,
 				   drop_cache);
 	btrfs_free_path(path);
 	return ret;
@@ -1645,6 +1648,324 @@ static int btrfs_file_mmap(struct file	*filp, struct vm_area_struct *vma)
 	return 0;
 }
 
+static int hole_mergeable(struct inode *inode, struct extent_buffer *leaf,
+			  int slot, u64 start, u64 end)
+{
+	struct btrfs_file_extent_item *fi;
+	struct btrfs_key key;
+
+	if (slot < 0 || slot >= btrfs_header_nritems(leaf))
+		return 0;
+
+	btrfs_item_key_to_cpu(leaf, &key, slot);
+	if (key.objectid != btrfs_ino(inode) ||
+	    key.type != BTRFS_EXTENT_DATA_KEY)
+		return 0;
+
+	fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
+
+	if (btrfs_file_extent_type(leaf, fi) != BTRFS_FILE_EXTENT_REG)
+		return 0;
+
+	if (btrfs_file_extent_disk_bytenr(leaf, fi))
+		return 0;
+
+	if (key.offset == end)
+		return 1;
+	if (key.offset + btrfs_file_extent_num_bytes(leaf, fi) == start)
+		return 1;
+	return 0;
+}
+
+static int fill_holes(struct btrfs_trans_handle *trans, struct inode *inode,
+		      struct btrfs_path *path, u64 offset, u64 end)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct extent_buffer *leaf;
+	struct btrfs_file_extent_item *fi;
+	struct extent_map *hole_em;
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct btrfs_key key;
+	int ret;
+
+	key.objectid = btrfs_ino(inode);
+	key.type = BTRFS_EXTENT_DATA_KEY;
+	key.offset = offset;
+
+
+	ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
+	if (ret < 0)
+		return ret;
+	BUG_ON(!ret);
+
+	leaf = path->nodes[0];
+	if (hole_mergeable(inode, leaf, path->slots[0]-1, offset, end)) {
+		u64 num_bytes;
+
+		path->slots[0]--;
+		fi = btrfs_item_ptr(leaf, path->slots[0],
+				    struct btrfs_file_extent_item);
+		num_bytes = btrfs_file_extent_num_bytes(leaf, fi) +
+			end - offset;
+		btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
+		btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
+		btrfs_set_file_extent_offset(leaf, fi, 0);
+		btrfs_mark_buffer_dirty(leaf);
+		goto out;
+	}
+
+	if (hole_mergeable(inode, leaf, path->slots[0]+1, offset, end)) {
+		u64 num_bytes;
+
+		path->slots[0]++;
+		key.offset = offset;
+		btrfs_set_item_key_safe(trans, root, path, &key);
+		fi = btrfs_item_ptr(leaf, path->slots[0],
+				    struct btrfs_file_extent_item);
+		num_bytes = btrfs_file_extent_num_bytes(leaf, fi) + end -
+			offset;
+		btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
+		btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
+		btrfs_set_file_extent_offset(leaf, fi, 0);
+		btrfs_mark_buffer_dirty(leaf);
+		goto out;
+	}
+	btrfs_release_path(path);
+
+	ret = btrfs_insert_file_extent(trans, root, btrfs_ino(inode), offset,
+				       0, 0, end - offset, 0, end - offset,
+				       0, 0, 0);
+	if (ret)
+		return ret;
+
+out:
+	btrfs_release_path(path);
+
+	hole_em = alloc_extent_map();
+	if (!hole_em) {
+		btrfs_drop_extent_cache(inode, offset, end - 1, 0);
+		set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+			&BTRFS_I(inode)->runtime_flags);
+	} else {
+		hole_em->start = offset;
+		hole_em->len = end - offset;
+		hole_em->orig_start = offset;
+
+		hole_em->block_start = EXTENT_MAP_HOLE;
+		hole_em->block_len = 0;
+		hole_em->bdev = root->fs_info->fs_devices->latest_bdev;
+		hole_em->compress_type = BTRFS_COMPRESS_NONE;
+		hole_em->generation = trans->transid;
+
+		do {
+			btrfs_drop_extent_cache(inode, offset, end - 1, 0);
+			write_lock(&em_tree->lock);
+			ret = add_extent_mapping(em_tree, hole_em);
+			if (!ret)
+				list_move(&hole_em->list,
+					  &em_tree->modified_extents);
+			write_unlock(&em_tree->lock);
+		} while (ret == -EEXIST);
+		free_extent_map(hole_em);
+		if (ret)
+			set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+				&BTRFS_I(inode)->runtime_flags);
+	}
+
+	return 0;
+}
+
+static int btrfs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct extent_state *cached_state = NULL;
+	struct btrfs_path *path;
+	struct btrfs_block_rsv *rsv;
+	struct btrfs_trans_handle *trans;
+	u64 mask = BTRFS_I(inode)->root->sectorsize - 1;
+	u64 lockstart = (offset + mask) & ~mask;
+	u64 lockend = ((offset + len) & ~mask) - 1;
+	u64 cur_offset = lockstart;
+	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
+	u64 drop_end;
+	unsigned long nr;
+	int ret = 0;
+	int err = 0;
+	bool same_page = (offset >> PAGE_CACHE_SHIFT) ==
+		((offset + len) >> PAGE_CACHE_SHIFT);
+
+	btrfs_wait_ordered_range(inode, offset, len);
+
+	mutex_lock(&inode->i_mutex);
+	if (offset >= inode->i_size) {
+		mutex_unlock(&inode->i_mutex);
+		return 0;
+	}
+
+	/*
+	 * Only do this if we are in the same page and we aren't doing the
+	 * entire page.
+	 */
+	if (same_page && len < PAGE_CACHE_SIZE) {
+		ret = btrfs_truncate_page(inode, offset, len, 0);
+		mutex_unlock(&inode->i_mutex);
+		return ret;
+	}
+
+	/* zero back part of the first page */
+	ret = btrfs_truncate_page(inode, offset, 0, 0);
+	if (ret) {
+		mutex_unlock(&inode->i_mutex);
+		return ret;
+	}
+
+	/* zero the front end of the last page */
+	ret = btrfs_truncate_page(inode, offset + len, 0, 1);
+	if (ret) {
+		mutex_unlock(&inode->i_mutex);
+		return ret;
+	}
+
+	if (lockend < lockstart) {
+		mutex_unlock(&inode->i_mutex);
+		return 0;
+	}
+
+	while (1) {
+		struct btrfs_ordered_extent *ordered;
+
+		truncate_pagecache_range(inode, lockstart, lockend);
+
+		lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend,
+				 0, &cached_state);
+		ordered = btrfs_lookup_first_ordered_extent(inode, lockend);
+
+		/*
+		 * We need to make sure we have no ordered extents in this range
+		 * and nobody raced in and read a page in this range, if we did
+		 * we need to try again.
+		 */
+		if ((!ordered ||
+		    (ordered->file_offset + ordered->len < lockstart ||
+		     ordered->file_offset > lockend)) &&
+		     !test_range_bit(&BTRFS_I(inode)->io_tree, lockstart,
+				     lockend, EXTENT_UPTODATE, 0,
+				     cached_state)) {
+			if (ordered)
+				btrfs_put_ordered_extent(ordered);
+			break;
+		}
+		if (ordered)
+			btrfs_put_ordered_extent(ordered);
+		unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart,
+				     lockend, &cached_state, GFP_NOFS);
+		btrfs_wait_ordered_range(inode, lockstart,
+					 lockend - lockstart + 1);
+	}
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	rsv = btrfs_alloc_block_rsv(root);
+	if (!rsv) {
+		ret = -ENOMEM;
+		goto out_free;
+	}
+	rsv->size = btrfs_calc_trunc_metadata_size(root, 1);
+	rsv->failfast = 1;
+
+	/*
+	 * 1 - update the inode
+	 * 1 - removing the extents in the range
+	 * 1 - adding the hole extent
+	 */
+	trans = btrfs_start_transaction(root, 3);
+	if (IS_ERR(trans)) {
+		err = PTR_ERR(trans);
+		goto out_free;
+	}
+
+	ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv, rsv,
+				      min_size);
+	BUG_ON(ret);
+	trans->block_rsv = rsv;
+
+	while (cur_offset < lockend) {
+		ret = __btrfs_drop_extents(trans, root, inode, path,
+					   cur_offset, lockend + 1,
+					   &drop_end, 1);
+		if (ret != -ENOSPC)
+			break;
+
+		trans->block_rsv = &root->fs_info->trans_block_rsv;
+
+		ret = fill_holes(trans, inode, path, cur_offset, drop_end);
+		if (ret) {
+			err = ret;
+			break;
+		}
+
+		cur_offset = drop_end;
+
+		ret = btrfs_update_inode(trans, root, inode);
+		if (ret) {
+			err = ret;
+			break;
+		}
+
+		nr = trans->blocks_used;
+		btrfs_end_transaction(trans, root);
+		btrfs_btree_balance_dirty(root, nr);
+
+		trans = btrfs_start_transaction(root, 3);
+		if (IS_ERR(trans)) {
+			ret = PTR_ERR(trans);
+			trans = NULL;
+			break;
+		}
+
+		ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv,
+					      rsv, min_size);
+		BUG_ON(ret);	/* shouldn't happen */
+		trans->block_rsv = rsv;
+	}
+
+	if (ret) {
+		err = ret;
+		goto out_trans;
+	}
+
+	trans->block_rsv = &root->fs_info->trans_block_rsv;
+	ret = fill_holes(trans, inode, path, cur_offset, drop_end);
+	if (ret) {
+		err = ret;
+		goto out_trans;
+	}
+
+out_trans:
+	if (!trans)
+		goto out_free;
+
+	trans->block_rsv = &root->fs_info->trans_block_rsv;
+	ret = btrfs_update_inode(trans, root, inode);
+	nr = trans->blocks_used;
+	btrfs_end_transaction(trans, root);
+	btrfs_btree_balance_dirty(root, nr);
+out_free:
+	btrfs_free_path(path);
+	btrfs_free_block_rsv(root, rsv);
+out:
+	unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
+			     &cached_state, GFP_NOFS);
+	mutex_unlock(&inode->i_mutex);
+	if (ret && !err)
+		err = ret;
+	return err;
+}
+
 static long btrfs_fallocate(struct file *file, int mode,
 			    loff_t offset, loff_t len)
 {
@@ -1663,10 +1984,13 @@ static long btrfs_fallocate(struct file *file, int mode,
 	alloc_start = offset & ~mask;
 	alloc_end =  (offset + len + mask) & ~mask;
 
-	/* We only support the FALLOC_FL_KEEP_SIZE mode */
+	/* Make sure we aren't being give some crap mode */
-	if (mode & ~FALLOC_FL_KEEP_SIZE)
+	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
 		return -EOPNOTSUPP;
 
+	if (mode & FALLOC_FL_PUNCH_HOLE)
+		return btrfs_punch_hole(inode, offset, len);
+
 	/*
 	 * Make sure we have enough space before we do the
 	 * allocation.

fs/btrfs/inode.c

@@ -3475,12 +3475,20 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
 }
 
 /*
- * taken from block_truncate_page, but does cow as it zeros out
+ * btrfs_truncate_page - read, zero a chunk and write a page
- * any bytes left in the last page in the file.
+ * @inode - inode that we're zeroing
+ * @from - the offset to start zeroing
+ * @len - the length to zero, 0 to zero the entire range respective to the
+ *	offset
+ * @front - zero up to the offset instead of from the offset on
+ *
+ * This will find the page for the "from" offset and cow the page and zero the
+ * part we want to zero.  This is used with truncate and hole punching.
  */
-static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
+int btrfs_truncate_page(struct inode *inode, loff_t from, loff_t len,
+			int front)
 {
-	struct inode *inode = mapping->host;
+	struct address_space *mapping = inode->i_mapping;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 	struct btrfs_ordered_extent *ordered;
@@ -3495,7 +3503,8 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 	u64 page_start;
 	u64 page_end;
 
-	if ((offset & (blocksize - 1)) == 0)
+	if ((offset & (blocksize - 1)) == 0 &&
+	    (!len || ((len & (blocksize - 1)) == 0)))
 		goto out;
 	ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
 	if (ret)
@@ -3555,8 +3564,13 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 
 	ret = 0;
 	if (offset != PAGE_CACHE_SIZE) {
+		if (!len)
+			len = PAGE_CACHE_SIZE - offset;
 		kaddr = kmap(page);
-		memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
+		if (front)
+			memset(kaddr, 0, offset);
+		else
+			memset(kaddr + offset, 0, len);
 		flush_dcache_page(page);
 		kunmap(page);
 	}
@@ -6796,7 +6810,7 @@ static int btrfs_truncate(struct inode *inode)
 	u64 mask = root->sectorsize - 1;
 	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
 
-	ret = btrfs_truncate_page(inode->i_mapping, inode->i_size);
+	ret = btrfs_truncate_page(inode, inode->i_size, 0, 0);
 	if (ret)
 		return ret;
 

fs/btrfs/tree-log.c

@@ -2842,7 +2842,7 @@ static int log_one_extent(struct btrfs_trans_handle *trans,
 
 	if (BTRFS_I(inode)->logged_trans == trans->transid) {
 		ret = __btrfs_drop_extents(trans, log, inode, dst_path, start,
-					   start + len, 0);
+					   start + len, NULL, 0);
 		if (ret)
 			return ret;
 	}