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>
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 2559f5d..10129cc 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -346,6 +346,8 @@
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 @@
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);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index ebb2db6..eebb4fb 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -223,7 +223,8 @@
} 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 @@
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 @@
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;
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index a0dff34..2b92f10 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -1131,7 +1131,8 @@
}
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 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 @@
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 @@
write_unlock_irq(&tree->lock);
return total_bytes;
}
-
/*
* helper function to lock both pages and extents in the tree.
* pages must be locked first.
diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h
index 6e572d3..ea60f54 100644
--- a/fs/btrfs/extent_map.h
+++ b/fs/btrfs/extent_map.h
@@ -115,7 +115,8 @@
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);
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 461b096..71dc2d3 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -307,6 +307,7 @@
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 @@
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 @@
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 @@
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 @@
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]);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 1c61c85..a9f5d6d 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -72,6 +72,26 @@
[S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
};
+int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
+ int for_del)
+{
+ u64 total = btrfs_super_total_bytes(&root->fs_info->super_copy);
+ u64 used = btrfs_super_bytes_used(&root->fs_info->super_copy);
+ u64 thresh;
+ int ret = 0;
+
+ if (for_del)
+ thresh = (total * 90) / 100;
+ else
+ thresh = (total * 85) / 100;
+
+ spin_lock(&root->fs_info->delalloc_lock);
+ if (used + root->fs_info->delalloc_bytes + num_required > thresh)
+ ret = -ENOSPC;
+ spin_unlock(&root->fs_info->delalloc_lock);
+ return ret;
+}
+
static int cow_file_range(struct inode *inode, u64 start, u64 end)
{
struct btrfs_root *root = BTRFS_I(inode)->root;
@@ -124,6 +144,7 @@
u64 extent_end;
u64 bytenr;
u64 cow_end;
+ u64 loops = 0;
struct btrfs_root *root = BTRFS_I(inode)->root;
struct extent_buffer *leaf;
int found_type;
@@ -169,6 +190,9 @@
btrfs_file_extent_num_bytes(leaf, item);
err = 0;
+ if (loops && start != extent_start)
+ goto not_found;
+
if (start < extent_start || start >= extent_end)
goto not_found;
@@ -191,6 +215,7 @@
return 0;
}
btrfs_release_path(root, path);
+ loops++;
goto again;
not_found:
@@ -202,6 +227,7 @@
static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
{
struct btrfs_root *root = BTRFS_I(inode)->root;
+ u64 num_bytes;
int ret;
mutex_lock(&root->fs_info->fs_mutex);
@@ -209,6 +235,17 @@
ret = run_delalloc_nocow(inode, start, end);
else
ret = cow_file_range(inode, start, end);
+
+ spin_lock(&root->fs_info->delalloc_lock);
+ num_bytes = end + 1 - start;
+ if (root->fs_info->delalloc_bytes < num_bytes) {
+ printk("delalloc accounting error total %llu sub %llu\n",
+ root->fs_info->delalloc_bytes, num_bytes);
+ } else {
+ root->fs_info->delalloc_bytes -= num_bytes;
+ }
+ spin_unlock(&root->fs_info->delalloc_lock);
+
mutex_unlock(&root->fs_info->fs_mutex);
return ret;
}
@@ -547,10 +584,15 @@
struct btrfs_root *root;
struct btrfs_trans_handle *trans;
int ret;
- unsigned long nr;
+ unsigned long nr = 0;
root = BTRFS_I(dir)->root;
mutex_lock(&root->fs_info->fs_mutex);
+
+ ret = btrfs_check_free_space(root, 1, 1);
+ if (ret)
+ goto fail;
+
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, dir);
@@ -558,25 +600,29 @@
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
+fail:
mutex_unlock(&root->fs_info->fs_mutex);
btrfs_btree_balance_dirty(root, nr);
-
return ret;
}
static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
{
struct inode *inode = dentry->d_inode;
- int err;
+ int err = 0;
int ret;
struct btrfs_root *root = BTRFS_I(dir)->root;
struct btrfs_trans_handle *trans;
- unsigned long nr;
+ unsigned long nr = 0;
if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
return -ENOTEMPTY;
mutex_lock(&root->fs_info->fs_mutex);
+ ret = btrfs_check_free_space(root, 1, 1);
+ if (ret)
+ goto fail;
+
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, dir);
@@ -588,6 +634,7 @@
nr = trans->blocks_used;
ret = btrfs_end_transaction(trans, root);
+fail:
mutex_unlock(&root->fs_info->fs_mutex);
btrfs_btree_balance_dirty(root, nr);
@@ -792,17 +839,29 @@
size_t zero_start)
{
char *kaddr;
- int ret = 0;
struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
+ u64 existing_delalloc;
+ u64 delalloc_start;
+ int ret = 0;
WARN_ON(!PageLocked(page));
set_page_extent_mapped(page);
lock_extent(em_tree, page_start, page_end, GFP_NOFS);
+ delalloc_start = page_start;
+ existing_delalloc = count_range_bits(&BTRFS_I(inode)->extent_tree,
+ &delalloc_start, page_end,
+ PAGE_CACHE_SIZE, EXTENT_DELALLOC);
set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start,
page_end, GFP_NOFS);
+
+ spin_lock(&root->fs_info->delalloc_lock);
+ root->fs_info->delalloc_bytes += PAGE_CACHE_SIZE - existing_delalloc;
+ spin_unlock(&root->fs_info->delalloc_lock);
+
if (zero_start != PAGE_CACHE_SIZE) {
kaddr = kmap(page);
memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
@@ -881,6 +940,12 @@
if (attr->ia_size <= pos)
goto out;
+ mutex_lock(&root->fs_info->fs_mutex);
+ err = btrfs_check_free_space(root, 1, 0);
+ mutex_unlock(&root->fs_info->fs_mutex);
+ if (err)
+ goto fail;
+
btrfs_truncate_page(inode->i_mapping, inode->i_size);
lock_extent(em_tree, pos, block_end, GFP_NOFS);
@@ -906,7 +971,7 @@
}
out:
err = inode_setattr(inode, attr);
-
+fail:
return err;
}
void btrfs_delete_inode(struct inode *inode)
@@ -1440,16 +1505,20 @@
{
struct btrfs_trans_handle *trans;
struct btrfs_root *root = BTRFS_I(dir)->root;
- struct inode *inode;
+ struct inode *inode = NULL;
int err;
int drop_inode = 0;
u64 objectid;
- unsigned long nr;
+ unsigned long nr = 0;
if (!new_valid_dev(rdev))
return -EINVAL;
mutex_lock(&root->fs_info->fs_mutex);
+ err = btrfs_check_free_space(root, 1, 0);
+ if (err)
+ goto fail;
+
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, dir);
@@ -1480,6 +1549,7 @@
out_unlock:
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
+fail:
mutex_unlock(&root->fs_info->fs_mutex);
if (drop_inode) {
@@ -1495,13 +1565,16 @@
{
struct btrfs_trans_handle *trans;
struct btrfs_root *root = BTRFS_I(dir)->root;
- struct inode *inode;
+ struct inode *inode = NULL;
int err;
int drop_inode = 0;
- unsigned long nr;
+ unsigned long nr = 0;
u64 objectid;
mutex_lock(&root->fs_info->fs_mutex);
+ err = btrfs_check_free_space(root, 1, 0);
+ if (err)
+ goto fail;
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, dir);
@@ -1535,6 +1608,7 @@
out_unlock:
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
+fail:
mutex_unlock(&root->fs_info->fs_mutex);
if (drop_inode) {
@@ -1551,7 +1625,7 @@
struct btrfs_trans_handle *trans;
struct btrfs_root *root = BTRFS_I(dir)->root;
struct inode *inode = old_dentry->d_inode;
- unsigned long nr;
+ unsigned long nr = 0;
int err;
int drop_inode = 0;
@@ -1564,6 +1638,9 @@
inc_nlink(inode);
#endif
mutex_lock(&root->fs_info->fs_mutex);
+ err = btrfs_check_free_space(root, 1, 0);
+ if (err)
+ goto fail;
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, dir);
@@ -1582,6 +1659,7 @@
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
+fail:
mutex_unlock(&root->fs_info->fs_mutex);
if (drop_inode) {
@@ -1603,6 +1681,10 @@
unsigned long nr = 1;
mutex_lock(&root->fs_info->fs_mutex);
+ err = btrfs_check_free_space(root, 1, 0);
+ if (err)
+ goto out_unlock;
+
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, dir);
@@ -1869,6 +1951,15 @@
static int btrfs_prepare_write(struct file *file, struct page *page,
unsigned from, unsigned to)
{
+ struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
+ int err;
+
+ mutex_lock(&root->fs_info->fs_mutex);
+ err = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
+ mutex_lock(&root->fs_info->fs_mutex);
+ if (err)
+ return -ENOSPC;
+
return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
page->mapping->host, page, from, to,
btrfs_get_extent);
@@ -1880,6 +1971,7 @@
tree = &BTRFS_I(page->mapping->host)->extent_tree;
return extent_read_full_page(tree, page, btrfs_get_extent);
}
+
static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
{
struct extent_map_tree *tree;
@@ -1954,11 +2046,20 @@
int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
{
struct inode *inode = fdentry(vma->vm_file)->d_inode;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
unsigned long end;
loff_t size;
- int ret = -EINVAL;
+ int ret;
u64 page_start;
+ mutex_lock(&root->fs_info->fs_mutex);
+ ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
+ mutex_lock(&root->fs_info->fs_mutex);
+ if (ret)
+ goto out;
+
+ ret = -EINVAL;
+
down_read(&BTRFS_I(inode)->root->snap_sem);
lock_page(page);
wait_on_page_writeback(page);
@@ -1982,6 +2083,7 @@
out_unlock:
up_read(&BTRFS_I(inode)->root->snap_sem);
unlock_page(page);
+out:
return ret;
}
@@ -2046,6 +2148,10 @@
unsigned long nr = 1;
mutex_lock(&root->fs_info->fs_mutex);
+ ret = btrfs_check_free_space(root, 1, 0);
+ if (ret)
+ goto fail_commit;
+
trans = btrfs_start_transaction(root, 1);
BUG_ON(!trans);
@@ -2162,7 +2268,7 @@
int ret;
int err;
u64 objectid;
- unsigned long nr;
+ unsigned long nr = 0;
if (!root->ref_cows)
return -EINVAL;
@@ -2172,6 +2278,10 @@
thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb);
mutex_lock(&root->fs_info->fs_mutex);
+ ret = btrfs_check_free_space(root, 1, 0);
+ if (ret)
+ goto fail_unlock;
+
trans = btrfs_start_transaction(root, 1);
BUG_ON(!trans);
@@ -2229,7 +2339,7 @@
if (err && !ret)
ret = err;
-
+fail_unlock:
mutex_unlock(&root->fs_info->fs_mutex);
up_write(&root->snap_sem);
btrfs_btree_balance_dirty(root, nr);
@@ -2255,6 +2365,7 @@
int btrfs_defrag_file(struct file *file) {
struct inode *inode = fdentry(file)->d_inode;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
struct page *page;
unsigned long last_index;
@@ -2262,6 +2373,13 @@
u64 page_start;
u64 page_end;
unsigned long i;
+ int ret;
+
+ mutex_lock(&root->fs_info->fs_mutex);
+ ret = btrfs_check_free_space(root, inode->i_size, 0);
+ mutex_unlock(&root->fs_info->fs_mutex);
+ if (ret)
+ return -ENOSPC;
mutex_lock(&inode->i_mutex);
last_index = inode->i_size >> PAGE_CACHE_SHIFT;
@@ -2522,6 +2640,10 @@
}
mutex_lock(&root->fs_info->fs_mutex);
+ ret = btrfs_check_free_space(root, 1, 0);
+ if (ret)
+ goto out_unlock;
+
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, new_dir);
@@ -2553,6 +2675,7 @@
out_fail:
btrfs_free_path(path);
btrfs_end_transaction(trans, root);
+out_unlock:
mutex_unlock(&root->fs_info->fs_mutex);
return ret;
}
@@ -2564,7 +2687,7 @@
struct btrfs_root *root = BTRFS_I(dir)->root;
struct btrfs_path *path;
struct btrfs_key key;
- struct inode *inode;
+ struct inode *inode = NULL;
int err;
int drop_inode = 0;
u64 objectid;
@@ -2573,12 +2696,17 @@
unsigned long ptr;
struct btrfs_file_extent_item *ei;
struct extent_buffer *leaf;
- unsigned long nr;
+ unsigned long nr = 0;
name_len = strlen(symname) + 1;
if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
return -ENAMETOOLONG;
+
mutex_lock(&root->fs_info->fs_mutex);
+ err = btrfs_check_free_space(root, 1, 0);
+ if (err)
+ goto out_fail;
+
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, dir);
@@ -2645,6 +2773,7 @@
out_unlock:
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
+out_fail:
mutex_unlock(&root->fs_info->fs_mutex);
if (drop_inode) {
inode_dec_link_count(inode);
