Btrfs: Implement basic support for -ENOSPC

This is intended to prevent accidentally filling the drive.  A determined
user can still make things oops.

It includes some accounting of the current bytes under delayed allocation,
but this will change as things get optimized
Signed-off-by: Chris Mason <chris.mason@oracle.com>

fs/btrfs/ctree.h

@@ -346,6 +346,8 @@ struct btrfs_fs_info {
 	int closing;
 
 	u64 total_pinned;
+	spinlock_t delalloc_lock;
+	u64 delalloc_bytes;
 };
 /*
  * in ram representation of the tree.  extent_root is used for all allocations
@@ -1115,6 +1117,8 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root, struct btrfs_path *path,
 			u64 isize);
 /* inode.c */
+int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
+			   int for_del);
 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page);
 int btrfs_readpage(struct file *file, struct page *page);
 void btrfs_delete_inode(struct inode *inode);

fs/btrfs/disk-io.c

@@ -223,7 +223,8 @@ static int btree_writepages(struct address_space *mapping,
 		} else {
 			thresh = 8 * 1024 * 1024;
 		}
-		num_dirty = count_range_bits(tree, &start, thresh, EXTENT_DIRTY);
+		num_dirty = count_range_bits(tree, &start, (u64)-1,
+					     thresh, EXTENT_DIRTY);
 		if (num_dirty < thresh) {
 			return 0;
 		}
@@ -559,6 +560,7 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	INIT_LIST_HEAD(&fs_info->dead_roots);
 	INIT_LIST_HEAD(&fs_info->hashers);
 	spin_lock_init(&fs_info->hash_lock);
+	spin_lock_init(&fs_info->delalloc_lock);
 
 	memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj));
 	init_completion(&fs_info->kobj_unregister);
@@ -570,6 +572,7 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	fs_info->sb = sb;
 	fs_info->mount_opt = 0;
 	fs_info->max_extent = (u64)-1;
+	fs_info->delalloc_bytes = 0;
 	fs_info->btree_inode = new_inode(sb);
 	fs_info->btree_inode->i_ino = 1;
 	fs_info->btree_inode->i_nlink = 1;

fs/btrfs/extent_map.c

@@ -1131,7 +1131,8 @@ u64 find_lock_delalloc_range(struct extent_map_tree *tree,
 }
 
 u64 count_range_bits(struct extent_map_tree *tree,
-		     u64 *start, u64 max_bytes, unsigned long bits)
+		     u64 *start, u64 search_end, u64 max_bytes,
+		     unsigned long bits)
 {
 	struct rb_node *node;
 	struct extent_state *state;
@@ -1139,9 +1140,14 @@ u64 count_range_bits(struct extent_map_tree *tree,
 	u64 total_bytes = 0;
 	int found = 0;
 
+	if (search_end <= cur_start) {
+		printk("search_end %Lu start %Lu\n", search_end, cur_start);
+		WARN_ON(1);
+		return 0;
+	}
+
 	write_lock_irq(&tree->lock);
-	if (bits == EXTENT_DIRTY) {
-		*start = 0;
+	if (cur_start == 0 && bits == EXTENT_DIRTY) {
 		total_bytes = tree->dirty_bytes;
 		goto out;
 	}
@@ -1156,8 +1162,11 @@ u64 count_range_bits(struct extent_map_tree *tree,
 
 	while(1) {
 		state = rb_entry(node, struct extent_state, rb_node);
-		if ((state->state & bits)) {
-			total_bytes += state->end - state->start + 1;
+		if (state->start > search_end)
+			break;
+		if (state->end >= cur_start && (state->state & bits)) {
+			total_bytes += min(search_end, state->end) + 1 -
+				       max(cur_start, state->start);
 			if (total_bytes >= max_bytes)
 				break;
 			if (!found) {
@@ -1173,7 +1182,6 @@ u64 count_range_bits(struct extent_map_tree *tree,
 	write_unlock_irq(&tree->lock);
 	return total_bytes;
 }
-
 /*
  * helper function to lock both pages and extents in the tree.
  * pages must be locked first.

fs/btrfs/extent_map.h

@@ -115,7 +115,8 @@ int __init extent_map_init(void);
 void extent_map_exit(void);
 
 u64 count_range_bits(struct extent_map_tree *tree,
-		     u64 *start, u64 max_bytes, unsigned long bits);
+		     u64 *start, u64 search_end,
+		     u64 max_bytes, unsigned long bits);
 
 int test_range_bit(struct extent_map_tree *tree, u64 start, u64 end,
 		   int bits, int filled);

fs/btrfs/file.c

@@ -307,6 +307,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	    inline_size > 32768 ||
 	    inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
 		u64 last_end;
+		u64 existing_delalloc = 0;
 
 		for (i = 0; i < num_pages; i++) {
 			struct page *p = pages[i];
@@ -316,8 +317,19 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 		last_end = (u64)(pages[num_pages -1]->index) <<
 				PAGE_CACHE_SHIFT;
 		last_end += PAGE_CACHE_SIZE - 1;
+		if (start_pos < isize) {
+			u64 delalloc_start = start_pos;
+			existing_delalloc = count_range_bits(em_tree,
+					     &delalloc_start,
+					     end_of_last_block, (u64)-1,
+					     EXTENT_DELALLOC);
+		}
 		set_extent_delalloc(em_tree, start_pos, end_of_last_block,
 				 GFP_NOFS);
+		spin_lock(&root->fs_info->delalloc_lock);
+		root->fs_info->delalloc_bytes += (end_of_last_block + 1 -
+					  start_pos) - existing_delalloc;
+		spin_unlock(&root->fs_info->delalloc_lock);
 	} else {
 		u64 aligned_end;
 		/* step one, delete the existing extents in this range */
@@ -708,12 +720,12 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 	current->backing_dev_info = inode->i_mapping->backing_dev_info;
 	err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
 	if (err)
-		goto out;
+		goto out_nolock;
 	if (count == 0)
-		goto out;
+		goto out_nolock;
 	err = remove_suid(fdentry(file));
 	if (err)
-		goto out;
+		goto out_nolock;
 	file_update_time(file);
 
 	pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL);
@@ -758,6 +770,13 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 
 		WARN_ON(num_pages > nrptrs);
 		memset(pages, 0, sizeof(pages));
+
+		mutex_lock(&root->fs_info->fs_mutex);
+		ret = btrfs_check_free_space(root, write_bytes, 0);
+		mutex_unlock(&root->fs_info->fs_mutex);
+		if (ret)
+			goto out;
+
 		ret = prepare_pages(root, file, pages, num_pages,
 				    pos, first_index, last_index,
 				    write_bytes);
@@ -787,8 +806,9 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 			btrfs_btree_balance_dirty(root, 1);
 		cond_resched();
 	}
-	mutex_unlock(&inode->i_mutex);
 out:
+	mutex_unlock(&inode->i_mutex);
+out_nolock:
 	kfree(pages);
 	if (pinned[0])
 		page_cache_release(pinned[0]);