Btrfs: fix enospc problems with delalloc
So I had this brilliant idea to use atomic counters for outstanding and reserved
extents, but this turned out to be a bad idea. Consider this where we have 1
outstanding extent and 1 reserved extent
Reserver Releaser
atomic_dec(outstanding) now 0
atomic_read(outstanding)+1 get 1
atomic_read(reserved) get 1
don't actually reserve anything because
they are the same
atomic_cmpxchg(reserved, 1, 0)
atomic_inc(outstanding)
atomic_add(0, reserved)
free reserved space for 1 extent
Then the reserver now has no actual space reserved for it, and when it goes to
finish the ordered IO it won't have enough space to do it's allocation and you
get those lovely warnings.
Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index 52d7eca..03dce3f 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -34,6 +34,9 @@
*/
struct btrfs_key location;
+ /* Lock for counters */
+ spinlock_t lock;
+
/* the extent_tree has caches of all the extent mappings to disk */
struct extent_map_tree extent_tree;
@@ -134,8 +137,8 @@
* items we think we'll end up using, and reserved_extents is the number
* of extent items we've reserved metadata for.
*/
- atomic_t outstanding_extents;
- atomic_t reserved_extents;
+ unsigned outstanding_extents;
+ unsigned reserved_extents;
/*
* ordered_data_close is set by truncate when a file that used
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 9f6f342..3063f21 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -2134,7 +2134,7 @@
/* extent-tree.c */
static inline u64 btrfs_calc_trans_metadata_size(struct btrfs_root *root,
- int num_items)
+ unsigned num_items)
{
return (root->leafsize + root->nodesize * (BTRFS_MAX_LEVEL - 1)) *
3 * num_items;
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 0a5bd67..d7031e7 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -3726,7 +3726,6 @@
if (commit_trans) {
if (trans)
return -EAGAIN;
-
trans = btrfs_join_transaction(root);
BUG_ON(IS_ERR(trans));
ret = btrfs_commit_transaction(trans, root);
@@ -3946,6 +3945,30 @@
return block_rsv_migrate_bytes(src_rsv, dst_rsv, num_bytes);
}
+static unsigned drop_outstanding_extent(struct inode *inode)
+{
+ unsigned dropped_extents = 0;
+
+ spin_lock(&BTRFS_I(inode)->lock);
+ BUG_ON(!BTRFS_I(inode)->outstanding_extents);
+ BTRFS_I(inode)->outstanding_extents--;
+
+ /*
+ * If we have more or the same amount of outsanding extents than we have
+ * reserved then we need to leave the reserved extents count alone.
+ */
+ if (BTRFS_I(inode)->outstanding_extents >=
+ BTRFS_I(inode)->reserved_extents)
+ goto out;
+
+ dropped_extents = BTRFS_I(inode)->reserved_extents -
+ BTRFS_I(inode)->outstanding_extents;
+ BTRFS_I(inode)->reserved_extents -= dropped_extents;
+out:
+ spin_unlock(&BTRFS_I(inode)->lock);
+ return dropped_extents;
+}
+
static u64 calc_csum_metadata_size(struct inode *inode, u64 num_bytes)
{
return num_bytes >>= 3;
@@ -3955,9 +3978,8 @@
{
struct btrfs_root *root = BTRFS_I(inode)->root;
struct btrfs_block_rsv *block_rsv = &root->fs_info->delalloc_block_rsv;
- u64 to_reserve;
- int nr_extents;
- int reserved_extents;
+ u64 to_reserve = 0;
+ unsigned nr_extents = 0;
int ret;
if (btrfs_transaction_in_commit(root->fs_info))
@@ -3965,24 +3987,31 @@
num_bytes = ALIGN(num_bytes, root->sectorsize);
- nr_extents = atomic_read(&BTRFS_I(inode)->outstanding_extents) + 1;
- reserved_extents = atomic_read(&BTRFS_I(inode)->reserved_extents);
+ spin_lock(&BTRFS_I(inode)->lock);
+ BTRFS_I(inode)->outstanding_extents++;
- if (nr_extents > reserved_extents) {
- nr_extents -= reserved_extents;
+ if (BTRFS_I(inode)->outstanding_extents >
+ BTRFS_I(inode)->reserved_extents) {
+ nr_extents = BTRFS_I(inode)->outstanding_extents -
+ BTRFS_I(inode)->reserved_extents;
+ BTRFS_I(inode)->reserved_extents += nr_extents;
+
to_reserve = btrfs_calc_trans_metadata_size(root, nr_extents);
- } else {
- nr_extents = 0;
- to_reserve = 0;
}
+ spin_unlock(&BTRFS_I(inode)->lock);
to_reserve += calc_csum_metadata_size(inode, num_bytes);
ret = reserve_metadata_bytes(NULL, root, block_rsv, to_reserve, 1);
- if (ret)
+ if (ret) {
+ unsigned dropped;
+ /*
+ * We don't need the return value since our reservation failed,
+ * we just need to clean up our counter.
+ */
+ dropped = drop_outstanding_extent(inode);
+ WARN_ON(dropped > 1);
return ret;
-
- atomic_add(nr_extents, &BTRFS_I(inode)->reserved_extents);
- atomic_inc(&BTRFS_I(inode)->outstanding_extents);
+ }
block_rsv_add_bytes(block_rsv, to_reserve, 1);
@@ -3992,36 +4021,15 @@
void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes)
{
struct btrfs_root *root = BTRFS_I(inode)->root;
- u64 to_free;
- int nr_extents;
- int reserved_extents;
+ u64 to_free = 0;
+ unsigned dropped;
num_bytes = ALIGN(num_bytes, root->sectorsize);
- atomic_dec(&BTRFS_I(inode)->outstanding_extents);
- WARN_ON(atomic_read(&BTRFS_I(inode)->outstanding_extents) < 0);
-
- reserved_extents = atomic_read(&BTRFS_I(inode)->reserved_extents);
- do {
- int old, new;
-
- nr_extents = atomic_read(&BTRFS_I(inode)->outstanding_extents);
- if (nr_extents >= reserved_extents) {
- nr_extents = 0;
- break;
- }
- old = reserved_extents;
- nr_extents = reserved_extents - nr_extents;
- new = reserved_extents - nr_extents;
- old = atomic_cmpxchg(&BTRFS_I(inode)->reserved_extents,
- reserved_extents, new);
- if (likely(old == reserved_extents))
- break;
- reserved_extents = old;
- } while (1);
+ dropped = drop_outstanding_extent(inode);
to_free = calc_csum_metadata_size(inode, num_bytes);
- if (nr_extents > 0)
- to_free += btrfs_calc_trans_metadata_size(root, nr_extents);
+ if (dropped > 0)
+ to_free += btrfs_calc_trans_metadata_size(root, dropped);
btrfs_block_rsv_release(root, &root->fs_info->delalloc_block_rsv,
to_free);
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index bd6bbb8..6e56a46 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -1239,9 +1239,11 @@
* managed to copy.
*/
if (num_pages > dirty_pages) {
- if (copied > 0)
- atomic_inc(
- &BTRFS_I(inode)->outstanding_extents);
+ if (copied > 0) {
+ spin_lock(&BTRFS_I(inode)->lock);
+ BTRFS_I(inode)->outstanding_extents++;
+ spin_unlock(&BTRFS_I(inode)->lock);
+ }
btrfs_delalloc_release_space(inode,
(num_pages - dirty_pages) <<
PAGE_CACHE_SHIFT);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 2bf4d2b..55d68ea 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1298,7 +1298,9 @@
if (!(orig->state & EXTENT_DELALLOC))
return 0;
- atomic_inc(&BTRFS_I(inode)->outstanding_extents);
+ spin_lock(&BTRFS_I(inode)->lock);
+ BTRFS_I(inode)->outstanding_extents++;
+ spin_unlock(&BTRFS_I(inode)->lock);
return 0;
}
@@ -1316,7 +1318,9 @@
if (!(other->state & EXTENT_DELALLOC))
return 0;
- atomic_dec(&BTRFS_I(inode)->outstanding_extents);
+ spin_lock(&BTRFS_I(inode)->lock);
+ BTRFS_I(inode)->outstanding_extents--;
+ spin_unlock(&BTRFS_I(inode)->lock);
return 0;
}
@@ -1339,10 +1343,13 @@
u64 len = state->end + 1 - state->start;
bool do_list = !is_free_space_inode(root, inode);
- if (*bits & EXTENT_FIRST_DELALLOC)
+ if (*bits & EXTENT_FIRST_DELALLOC) {
*bits &= ~EXTENT_FIRST_DELALLOC;
- else
- atomic_inc(&BTRFS_I(inode)->outstanding_extents);
+ } else {
+ spin_lock(&BTRFS_I(inode)->lock);
+ BTRFS_I(inode)->outstanding_extents++;
+ spin_unlock(&BTRFS_I(inode)->lock);
+ }
spin_lock(&root->fs_info->delalloc_lock);
BTRFS_I(inode)->delalloc_bytes += len;
@@ -1372,10 +1379,13 @@
u64 len = state->end + 1 - state->start;
bool do_list = !is_free_space_inode(root, inode);
- if (*bits & EXTENT_FIRST_DELALLOC)
+ if (*bits & EXTENT_FIRST_DELALLOC) {
*bits &= ~EXTENT_FIRST_DELALLOC;
- else if (!(*bits & EXTENT_DO_ACCOUNTING))
- atomic_dec(&BTRFS_I(inode)->outstanding_extents);
+ } else if (!(*bits & EXTENT_DO_ACCOUNTING)) {
+ spin_lock(&BTRFS_I(inode)->lock);
+ BTRFS_I(inode)->outstanding_extents--;
+ spin_unlock(&BTRFS_I(inode)->lock);
+ }
if (*bits & EXTENT_DO_ACCOUNTING)
btrfs_delalloc_release_metadata(inode, len);
@@ -6735,8 +6745,9 @@
ei->index_cnt = (u64)-1;
ei->last_unlink_trans = 0;
- atomic_set(&ei->outstanding_extents, 0);
- atomic_set(&ei->reserved_extents, 0);
+ spin_lock_init(&ei->lock);
+ ei->outstanding_extents = 0;
+ ei->reserved_extents = 0;
ei->ordered_data_close = 0;
ei->orphan_meta_reserved = 0;
@@ -6774,8 +6785,8 @@
WARN_ON(!list_empty(&inode->i_dentry));
WARN_ON(inode->i_data.nrpages);
- WARN_ON(atomic_read(&BTRFS_I(inode)->outstanding_extents));
- WARN_ON(atomic_read(&BTRFS_I(inode)->reserved_extents));
+ WARN_ON(BTRFS_I(inode)->outstanding_extents);
+ WARN_ON(BTRFS_I(inode)->reserved_extents);
/*
* This can happen where we create an inode, but somebody else also
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 09c9a8d..fd252ff 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -938,7 +938,9 @@
GFP_NOFS);
if (i_done != num_pages) {
- atomic_inc(&BTRFS_I(inode)->outstanding_extents);
+ spin_lock(&BTRFS_I(inode)->lock);
+ BTRFS_I(inode)->outstanding_extents++;
+ spin_unlock(&BTRFS_I(inode)->lock);
btrfs_delalloc_release_space(inode,
(num_pages - i_done) << PAGE_CACHE_SHIFT);
}