diff options
Diffstat (limited to 'fs/btrfs/delayed-ref.c')
| -rw-r--r-- | fs/btrfs/delayed-ref.c | 994 |
1 files changed, 635 insertions, 359 deletions
diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c index d6c01c096a4..6d16bea94e1 100644 --- a/fs/btrfs/delayed-ref.c +++ b/fs/btrfs/delayed-ref.c @@ -17,11 +17,16 @@ */ #include <linux/sched.h> +#include <linux/slab.h> #include <linux/sort.h> #include "ctree.h" #include "delayed-ref.h" #include "transaction.h" +struct kmem_cache *btrfs_delayed_ref_head_cachep; +struct kmem_cache *btrfs_delayed_tree_ref_cachep; +struct kmem_cache *btrfs_delayed_data_ref_cachep; +struct kmem_cache *btrfs_delayed_extent_op_cachep; /* * delayed back reference update tracking. For subvolume trees * we queue up extent allocations and backref maintenance for @@ -29,27 +34,100 @@ * add extents in the middle of btrfs_search_slot, and it allows * us to buffer up frequently modified backrefs in an rb tree instead * of hammering updates on the extent allocation tree. - * - * Right now this code is only used for reference counted trees, but - * the long term goal is to get rid of the similar code for delayed - * extent tree modifications. */ /* - * entries in the rb tree are ordered by the byte number of the extent - * and by the byte number of the parent block. + * compare two delayed tree backrefs with same bytenr and type + */ +static int comp_tree_refs(struct btrfs_delayed_tree_ref *ref2, + struct btrfs_delayed_tree_ref *ref1, int type) +{ + if (type == BTRFS_TREE_BLOCK_REF_KEY) { + if (ref1->root < ref2->root) + return -1; + if (ref1->root > ref2->root) + return 1; + } else { + if (ref1->parent < ref2->parent) + return -1; + if (ref1->parent > ref2->parent) + return 1; + } + return 0; +} + +/* + * compare two delayed data backrefs with same bytenr and type + */ +static int comp_data_refs(struct btrfs_delayed_data_ref *ref2, + struct btrfs_delayed_data_ref *ref1) +{ + if (ref1->node.type == BTRFS_EXTENT_DATA_REF_KEY) { + if (ref1->root < ref2->root) + return -1; + if (ref1->root > ref2->root) + return 1; + if (ref1->objectid < ref2->objectid) + return -1; + if (ref1->objectid > ref2->objectid) + return 1; + if (ref1->offset < ref2->offset) + return -1; + if (ref1->offset > ref2->offset) + return 1; + } else { + if (ref1->parent < ref2->parent) + return -1; + if (ref1->parent > ref2->parent) + return 1; + } + return 0; +} + +/* + * entries in the rb tree are ordered by the byte number of the extent, + * type of the delayed backrefs and content of delayed backrefs. */ -static int comp_entry(struct btrfs_delayed_ref_node *ref, - u64 bytenr, u64 parent) +static int comp_entry(struct btrfs_delayed_ref_node *ref2, + struct btrfs_delayed_ref_node *ref1, + bool compare_seq) { - if (bytenr < ref->bytenr) + if (ref1->bytenr < ref2->bytenr) return -1; - if (bytenr > ref->bytenr) + if (ref1->bytenr > ref2->bytenr) return 1; - if (parent < ref->parent) + if (ref1->is_head && ref2->is_head) + return 0; + if (ref2->is_head) + return -1; + if (ref1->is_head) + return 1; + if (ref1->type < ref2->type) return -1; - if (parent > ref->parent) + if (ref1->type > ref2->type) return 1; + if (ref1->no_quota > ref2->no_quota) + return 1; + if (ref1->no_quota < ref2->no_quota) + return -1; + /* merging of sequenced refs is not allowed */ + if (compare_seq) { + if (ref1->seq < ref2->seq) + return -1; + if (ref1->seq > ref2->seq) + return 1; + } + if (ref1->type == BTRFS_TREE_BLOCK_REF_KEY || + ref1->type == BTRFS_SHARED_BLOCK_REF_KEY) { + return comp_tree_refs(btrfs_delayed_node_to_tree_ref(ref2), + btrfs_delayed_node_to_tree_ref(ref1), + ref1->type); + } else if (ref1->type == BTRFS_EXTENT_DATA_REF_KEY || + ref1->type == BTRFS_SHARED_DATA_REF_KEY) { + return comp_data_refs(btrfs_delayed_node_to_data_ref(ref2), + btrfs_delayed_node_to_data_ref(ref1)); + } + BUG(); return 0; } @@ -59,20 +137,21 @@ static int comp_entry(struct btrfs_delayed_ref_node *ref, * inserted. */ static struct btrfs_delayed_ref_node *tree_insert(struct rb_root *root, - u64 bytenr, u64 parent, struct rb_node *node) { struct rb_node **p = &root->rb_node; struct rb_node *parent_node = NULL; struct btrfs_delayed_ref_node *entry; + struct btrfs_delayed_ref_node *ins; int cmp; + ins = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); while (*p) { parent_node = *p; entry = rb_entry(parent_node, struct btrfs_delayed_ref_node, rb_node); - cmp = comp_entry(entry, bytenr, parent); + cmp = comp_entry(entry, ins, 1); if (cmp < 0) p = &(*p)->rb_left; else if (cmp > 0) @@ -81,38 +160,77 @@ static struct btrfs_delayed_ref_node *tree_insert(struct rb_root *root, return entry; } - entry = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); + rb_link_node(node, parent_node, p); + rb_insert_color(node, root); + return NULL; +} + +/* insert a new ref to head ref rbtree */ +static struct btrfs_delayed_ref_head *htree_insert(struct rb_root *root, + struct rb_node *node) +{ + struct rb_node **p = &root->rb_node; + struct rb_node *parent_node = NULL; + struct btrfs_delayed_ref_head *entry; + struct btrfs_delayed_ref_head *ins; + u64 bytenr; + + ins = rb_entry(node, struct btrfs_delayed_ref_head, href_node); + bytenr = ins->node.bytenr; + while (*p) { + parent_node = *p; + entry = rb_entry(parent_node, struct btrfs_delayed_ref_head, + href_node); + + if (bytenr < entry->node.bytenr) + p = &(*p)->rb_left; + else if (bytenr > entry->node.bytenr) + p = &(*p)->rb_right; + else + return entry; + } + rb_link_node(node, parent_node, p); rb_insert_color(node, root); return NULL; } /* - * find an entry based on (bytenr,parent). This returns the delayed - * ref if it was able to find one, or NULL if nothing was in that spot + * find an head entry based on bytenr. This returns the delayed ref + * head if it was able to find one, or NULL if nothing was in that spot. + * If return_bigger is given, the next bigger entry is returned if no exact + * match is found. */ -static struct btrfs_delayed_ref_node *tree_search(struct rb_root *root, - u64 bytenr, u64 parent, - struct btrfs_delayed_ref_node **last) +static struct btrfs_delayed_ref_head * +find_ref_head(struct rb_root *root, u64 bytenr, + int return_bigger) { - struct rb_node *n = root->rb_node; - struct btrfs_delayed_ref_node *entry; - int cmp; + struct rb_node *n; + struct btrfs_delayed_ref_head *entry; + n = root->rb_node; + entry = NULL; while (n) { - entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node); - WARN_ON(!entry->in_tree); - if (last) - *last = entry; + entry = rb_entry(n, struct btrfs_delayed_ref_head, href_node); - cmp = comp_entry(entry, bytenr, parent); - if (cmp < 0) + if (bytenr < entry->node.bytenr) n = n->rb_left; - else if (cmp > 0) + else if (bytenr > entry->node.bytenr) n = n->rb_right; else return entry; } + if (entry && return_bigger) { + if (bytenr > entry->node.bytenr) { + n = rb_next(&entry->href_node); + if (!n) + n = rb_first(root); + entry = rb_entry(n, struct btrfs_delayed_ref_head, + href_node); + return entry; + } + return entry; + } return NULL; } @@ -140,188 +258,188 @@ int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans, return 0; } -int btrfs_find_ref_cluster(struct btrfs_trans_handle *trans, - struct list_head *cluster, u64 start) +static inline void drop_delayed_ref(struct btrfs_trans_handle *trans, + struct btrfs_delayed_ref_root *delayed_refs, + struct btrfs_delayed_ref_head *head, + struct btrfs_delayed_ref_node *ref) { - int count = 0; - struct btrfs_delayed_ref_root *delayed_refs; - struct rb_node *node; - struct btrfs_delayed_ref_node *ref; - struct btrfs_delayed_ref_head *head; - - delayed_refs = &trans->transaction->delayed_refs; - if (start == 0) { - node = rb_first(&delayed_refs->root); + if (btrfs_delayed_ref_is_head(ref)) { + head = btrfs_delayed_node_to_head(ref); + rb_erase(&head->href_node, &delayed_refs->href_root); } else { - ref = NULL; - tree_search(&delayed_refs->root, start, (u64)-1, &ref); - if (ref) { - struct btrfs_delayed_ref_node *tmp; - - node = rb_prev(&ref->rb_node); - while (node) { - tmp = rb_entry(node, - struct btrfs_delayed_ref_node, - rb_node); - if (tmp->bytenr < start) - break; - ref = tmp; - node = rb_prev(&ref->rb_node); - } - node = &ref->rb_node; - } else - node = rb_first(&delayed_refs->root); + assert_spin_locked(&head->lock); + rb_erase(&ref->rb_node, &head->ref_root); } -again: - while (node && count < 32) { - ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); - if (btrfs_delayed_ref_is_head(ref)) { - head = btrfs_delayed_node_to_head(ref); - if (list_empty(&head->cluster)) { - list_add_tail(&head->cluster, cluster); - delayed_refs->run_delayed_start = - head->node.bytenr; - count++; - - WARN_ON(delayed_refs->num_heads_ready == 0); - delayed_refs->num_heads_ready--; - } else if (count) { - /* the goal of the clustering is to find extents - * that are likely to end up in the same extent - * leaf on disk. So, we don't want them spread - * all over the tree. Stop now if we've hit - * a head that was already in use - */ - break; + ref->in_tree = 0; + btrfs_put_delayed_ref(ref); + atomic_dec(&delayed_refs->num_entries); + if (trans->delayed_ref_updates) + trans->delayed_ref_updates--; +} + +static int merge_ref(struct btrfs_trans_handle *trans, + struct btrfs_delayed_ref_root *delayed_refs, + struct btrfs_delayed_ref_head *head, + struct btrfs_delayed_ref_node *ref, u64 seq) +{ + struct rb_node *node; + int mod = 0; + int done = 0; + + node = rb_next(&ref->rb_node); + while (!done && node) { + struct btrfs_delayed_ref_node *next; + + next = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); + node = rb_next(node); + if (seq && next->seq >= seq) + break; + if (comp_entry(ref, next, 0)) + continue; + + if (ref->action == next->action) { + mod = next->ref_mod; + } else { + if (ref->ref_mod < next->ref_mod) { + struct btrfs_delayed_ref_node *tmp; + + tmp = ref; + ref = next; + next = tmp; + done = 1; } + mod = -next->ref_mod; + } + + drop_delayed_ref(trans, delayed_refs, head, next); + ref->ref_mod += mod; + if (ref->ref_mod == 0) { + drop_delayed_ref(trans, delayed_refs, head, ref); + done = 1; + } else { + /* + * You can't have multiples of the same ref on a tree + * block. + */ + WARN_ON(ref->type == BTRFS_TREE_BLOCK_REF_KEY || + ref->type == BTRFS_SHARED_BLOCK_REF_KEY); } - node = rb_next(node); - } - if (count) { - return 0; - } else if (start) { - /* - * we've gone to the end of the rbtree without finding any - * clusters. start from the beginning and try again - */ - start = 0; - node = rb_first(&delayed_refs->root); - goto again; } - return 1; + return done; } -/* - * This checks to see if there are any delayed refs in the - * btree for a given bytenr. It returns one if it finds any - * and zero otherwise. - * - * If it only finds a head node, it returns 0. - * - * The idea is to use this when deciding if you can safely delete an - * extent from the extent allocation tree. There may be a pending - * ref in the rbtree that adds or removes references, so as long as this - * returns one you need to leave the BTRFS_EXTENT_ITEM in the extent - * allocation tree. - */ -int btrfs_delayed_ref_pending(struct btrfs_trans_handle *trans, u64 bytenr) +void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans, + struct btrfs_fs_info *fs_info, + struct btrfs_delayed_ref_root *delayed_refs, + struct btrfs_delayed_ref_head *head) { - struct btrfs_delayed_ref_node *ref; - struct btrfs_delayed_ref_root *delayed_refs; - struct rb_node *prev_node; - int ret = 0; + struct rb_node *node; + u64 seq = 0; - delayed_refs = &trans->transaction->delayed_refs; - spin_lock(&delayed_refs->lock); + assert_spin_locked(&head->lock); + /* + * We don't have too much refs to merge in the case of delayed data + * refs. + */ + if (head->is_data) + return; + + spin_lock(&fs_info->tree_mod_seq_lock); + if (!list_empty(&fs_info->tree_mod_seq_list)) { + struct seq_list *elem; + + elem = list_first_entry(&fs_info->tree_mod_seq_list, + struct seq_list, list); + seq = elem->seq; + } + spin_unlock(&fs_info->tree_mod_seq_lock); + + node = rb_first(&head->ref_root); + while (node) { + struct btrfs_delayed_ref_node *ref; - ref = tree_search(&delayed_refs->root, bytenr, (u64)-1, NULL); - if (ref) { - prev_node = rb_prev(&ref->rb_node); - if (!prev_node) - goto out; - ref = rb_entry(prev_node, struct btrfs_delayed_ref_node, + ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); - if (ref->bytenr == bytenr) + /* We can't merge refs that are outside of our seq count */ + if (seq && ref->seq >= seq) + break; + if (merge_ref(trans, delayed_refs, head, ref, seq)) + node = rb_first(&head->ref_root); + else + node = rb_next(&ref->rb_node); + } +} + +int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, + struct btrfs_delayed_ref_root *delayed_refs, + u64 seq) +{ + struct seq_list *elem; + int ret = 0; + + spin_lock(&fs_info->tree_mod_seq_lock); + if (!list_empty(&fs_info->tree_mod_seq_list)) { + elem = list_first_entry(&fs_info->tree_mod_seq_list, + struct seq_list, list); + if (seq >= elem->seq) { + pr_debug("holding back delayed_ref %#x.%x, lowest is %#x.%x (%p)\n", + (u32)(seq >> 32), (u32)seq, + (u32)(elem->seq >> 32), (u32)elem->seq, + delayed_refs); ret = 1; + } } -out: - spin_unlock(&delayed_refs->lock); + + spin_unlock(&fs_info->tree_mod_seq_lock); return ret; } -/* - * helper function to lookup reference count - * - * the head node for delayed ref is used to store the sum of all the - * reference count modifications queued up in the rbtree. This way you - * can check to see what the reference count would be if all of the - * delayed refs are processed. - */ -int btrfs_lookup_extent_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 bytenr, - u64 num_bytes, u32 *refs) +struct btrfs_delayed_ref_head * +btrfs_select_ref_head(struct btrfs_trans_handle *trans) { - struct btrfs_delayed_ref_node *ref; - struct btrfs_delayed_ref_head *head; struct btrfs_delayed_ref_root *delayed_refs; - struct btrfs_path *path; - struct extent_buffer *leaf; - struct btrfs_extent_item *ei; - struct btrfs_key key; - u32 num_refs; - int ret; - - path = btrfs_alloc_path(); - if (!path) - return -ENOMEM; + struct btrfs_delayed_ref_head *head; + u64 start; + bool loop = false; - key.objectid = bytenr; - key.type = BTRFS_EXTENT_ITEM_KEY; - key.offset = num_bytes; delayed_refs = &trans->transaction->delayed_refs; + again: - ret = btrfs_search_slot(trans, root->fs_info->extent_root, - &key, path, 0, 0); - if (ret < 0) - goto out; - - if (ret == 0) { - leaf = path->nodes[0]; - ei = btrfs_item_ptr(leaf, path->slots[0], - struct btrfs_extent_item); - num_refs = btrfs_extent_refs(leaf, ei); - } else { - num_refs = 0; - ret = 0; + start = delayed_refs->run_delayed_start; + head = find_ref_head(&delayed_refs->href_root, start, 1); + if (!head && !loop) { + delayed_refs->run_delayed_start = 0; + start = 0; + loop = true; + head = find_ref_head(&delayed_refs->href_root, start, 1); + if (!head) + return NULL; + } else if (!head && loop) { + return NULL; } - spin_lock(&delayed_refs->lock); - ref = tree_search(&delayed_refs->root, bytenr, (u64)-1, NULL); - if (ref) { - head = btrfs_delayed_node_to_head(ref); - if (mutex_trylock(&head->mutex)) { - num_refs += ref->ref_mod; - mutex_unlock(&head->mutex); - *refs = num_refs; - goto out; + while (head->processing) { + struct rb_node *node; + + node = rb_next(&head->href_node); + if (!node) { + if (loop) + return NULL; + delayed_refs->run_delayed_start = 0; + start = 0; + loop = true; + goto again; } - - atomic_inc(&ref->refs); - spin_unlock(&delayed_refs->lock); - - btrfs_release_path(root->fs_info->extent_root, path); - - mutex_lock(&head->mutex); - mutex_unlock(&head->mutex); - btrfs_put_delayed_ref(ref); - goto again; - } else { - *refs = num_refs; + head = rb_entry(node, struct btrfs_delayed_ref_head, + href_node); } -out: - spin_unlock(&delayed_refs->lock); - btrfs_free_path(path); - return ret; + + head->processing = 1; + WARN_ON(delayed_refs->num_heads_ready == 0); + delayed_refs->num_heads_ready--; + delayed_refs->run_delayed_start = head->node.bytenr + + head->node.num_bytes; + return head; } /* @@ -335,19 +453,11 @@ out: static noinline void update_existing_ref(struct btrfs_trans_handle *trans, struct btrfs_delayed_ref_root *delayed_refs, + struct btrfs_delayed_ref_head *head, struct btrfs_delayed_ref_node *existing, struct btrfs_delayed_ref_node *update) { - struct btrfs_delayed_ref *existing_ref; - struct btrfs_delayed_ref *ref; - - existing_ref = btrfs_delayed_node_to_ref(existing); - ref = btrfs_delayed_node_to_ref(update); - - if (ref->pin) - existing_ref->pin = 1; - - if (ref->action != existing_ref->action) { + if (update->action != existing->action) { /* * this is effectively undoing either an add or a * drop. We decrement the ref_mod, and if it goes @@ -355,28 +465,14 @@ update_existing_ref(struct btrfs_trans_handle *trans, * every changing the extent allocation tree. */ existing->ref_mod--; - if (existing->ref_mod == 0) { - rb_erase(&existing->rb_node, - &delayed_refs->root); - existing->in_tree = 0; - btrfs_put_delayed_ref(existing); - delayed_refs->num_entries--; - if (trans->delayed_ref_updates) - trans->delayed_ref_updates--; - } + if (existing->ref_mod == 0) + drop_delayed_ref(trans, delayed_refs, head, existing); + else + WARN_ON(existing->type == BTRFS_TREE_BLOCK_REF_KEY || + existing->type == BTRFS_SHARED_BLOCK_REF_KEY); } else { - if (existing_ref->action == BTRFS_ADD_DELAYED_REF) { - /* if we're adding refs, make sure all the - * details match up. The extent could - * have been totally freed and reallocated - * by a different owner before the delayed - * ref entries were removed. - */ - existing_ref->owner_objectid = ref->owner_objectid; - existing_ref->generation = ref->generation; - existing_ref->root = ref->root; - existing->num_bytes = update->num_bytes; - } + WARN_ON(existing->type == BTRFS_TREE_BLOCK_REF_KEY || + existing->type == BTRFS_SHARED_BLOCK_REF_KEY); /* * the action on the existing ref matches * the action on the ref we're trying to add. @@ -401,7 +497,9 @@ update_existing_head_ref(struct btrfs_delayed_ref_node *existing, existing_ref = btrfs_delayed_node_to_head(existing); ref = btrfs_delayed_node_to_head(update); + BUG_ON(existing_ref->is_data != ref->is_data); + spin_lock(&existing_ref->lock); if (ref->must_insert_reserved) { /* if the extent was freed and then * reallocated before the delayed ref @@ -420,26 +518,45 @@ update_existing_head_ref(struct btrfs_delayed_ref_node *existing, } + if (ref->extent_op) { + if (!existing_ref->extent_op) { + existing_ref->extent_op = ref->extent_op; + } else { + if (ref->extent_op->update_key) { + memcpy(&existing_ref->extent_op->key, + &ref->extent_op->key, + sizeof(ref->extent_op->key)); + existing_ref->extent_op->update_key = 1; + } + if (ref->extent_op->update_flags) { + existing_ref->extent_op->flags_to_set |= + ref->extent_op->flags_to_set; + existing_ref->extent_op->update_flags = 1; + } + btrfs_free_delayed_extent_op(ref->extent_op); + } + } /* - * update the reference mod on the head to reflect this new operation + * update the reference mod on the head to reflect this new operation, + * only need the lock for this case cause we could be processing it + * currently, for refs we just added we know we're a-ok. */ existing->ref_mod += update->ref_mod; + spin_unlock(&existing_ref->lock); } /* - * helper function to actually insert a delayed ref into the rbtree. + * helper function to actually insert a head node into the rbtree. * this does all the dirty work in terms of maintaining the correct - * overall modification count in the head node and properly dealing - * with updating existing nodes as new modifications are queued. + * overall modification count. */ -static noinline int __btrfs_add_delayed_ref(struct btrfs_trans_handle *trans, - struct btrfs_delayed_ref_node *ref, - u64 bytenr, u64 num_bytes, u64 parent, u64 ref_root, - u64 ref_generation, u64 owner_objectid, int action, - int pin) +static noinline struct btrfs_delayed_ref_head * +add_delayed_ref_head(struct btrfs_fs_info *fs_info, + struct btrfs_trans_handle *trans, + struct btrfs_delayed_ref_node *ref, u64 bytenr, + u64 num_bytes, int action, int is_data) { - struct btrfs_delayed_ref_node *existing; - struct btrfs_delayed_ref *full_ref; + struct btrfs_delayed_ref_head *existing; struct btrfs_delayed_ref_head *head_ref = NULL; struct btrfs_delayed_ref_root *delayed_refs; int count_mod = 1; @@ -449,12 +566,10 @@ static noinline int __btrfs_add_delayed_ref(struct btrfs_trans_handle *trans, * the head node stores the sum of all the mods, so dropping a ref * should drop the sum in the head node by one. */ - if (parent == (u64)-1) { - if (action == BTRFS_DROP_DELAYED_REF) - count_mod = -1; - else if (action == BTRFS_UPDATE_DELAYED_HEAD) - count_mod = 0; - } + if (action == BTRFS_UPDATE_DELAYED_HEAD) + count_mod = 0; + else if (action == BTRFS_DROP_DELAYED_REF) + count_mod = -1; /* * BTRFS_ADD_DELAYED_EXTENT means that we need to update @@ -467,95 +582,213 @@ static noinline int __btrfs_add_delayed_ref(struct btrfs_trans_handle *trans, * Once we record must_insert_reserved, switch the action to * BTRFS_ADD_DELAYED_REF because other special casing is not required. */ - if (action == BTRFS_ADD_DELAYED_EXTENT) { + if (action == BTRFS_ADD_DELAYED_EXTENT) must_insert_reserved = 1; - action = BTRFS_ADD_DELAYED_REF; - } else { + else must_insert_reserved = 0; - } - delayed_refs = &trans->transaction->delayed_refs; /* first set the basic ref node struct up */ atomic_set(&ref->refs, 1); ref->bytenr = bytenr; - ref->parent = parent; + ref->num_bytes = num_bytes; ref->ref_mod = count_mod; + ref->type = 0; + ref->action = 0; + ref->is_head = 1; ref->in_tree = 1; - ref->num_bytes = num_bytes; + ref->seq = 0; - if (btrfs_delayed_ref_is_head(ref)) { - head_ref = btrfs_delayed_node_to_head(ref); - head_ref->must_insert_reserved = must_insert_reserved; - INIT_LIST_HEAD(&head_ref->cluster); - mutex_init(&head_ref->mutex); + head_ref = btrfs_delayed_node_to_head(ref); + head_ref->must_insert_reserved = must_insert_reserved; + head_ref->is_data = is_data; + head_ref->ref_root = RB_ROOT; + head_ref->processing = 0; + + spin_lock_init(&head_ref->lock); + mutex_init(&head_ref->mutex); + + trace_add_delayed_ref_head(ref, head_ref, action); + + existing = htree_insert(&delayed_refs->href_root, + &head_ref->href_node); + if (existing) { + update_existing_head_ref(&existing->node, ref); + /* + * we've updated the existing ref, free the newly + * allocated ref + */ + kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref); + head_ref = existing; } else { - full_ref = btrfs_delayed_node_to_ref(ref); - full_ref->root = ref_root; - full_ref->generation = ref_generation; - full_ref->owner_objectid = owner_objectid; - full_ref->pin = pin; - full_ref->action = action; + delayed_refs->num_heads++; + delayed_refs->num_heads_ready++; + atomic_inc(&delayed_refs->num_entries); + trans->delayed_ref_updates++; } + return head_ref; +} + +/* + * helper to insert a delayed tree ref into the rbtree. + */ +static noinline void +add_delayed_tree_ref(struct btrfs_fs_info *fs_info, + struct btrfs_trans_handle *trans, + struct btrfs_delayed_ref_head *head_ref, + struct btrfs_delayed_ref_node *ref, u64 bytenr, + u64 num_bytes, u64 parent, u64 ref_root, int level, + int action, int no_quota) +{ + struct btrfs_delayed_ref_node *existing; + struct btrfs_delayed_tree_ref *full_ref; + struct btrfs_delayed_ref_root *delayed_refs; + u64 seq = 0; - existing = tree_insert(&delayed_refs->root, bytenr, - parent, &ref->rb_node); + if (action == BTRFS_ADD_DELAYED_EXTENT) + action = BTRFS_ADD_DELAYED_REF; + + if (is_fstree(ref_root)) + seq = atomic64_read(&fs_info->tree_mod_seq); + delayed_refs = &trans->transaction->delayed_refs; + /* first set the basic ref node struct up */ + atomic_set(&ref->refs, 1); + ref->bytenr = bytenr; + ref->num_bytes = num_bytes; + ref->ref_mod = 1; + ref->action = action; + ref->is_head = 0; + ref->in_tree = 1; + ref->no_quota = no_quota; + ref->seq = seq; + + full_ref = btrfs_delayed_node_to_tree_ref(ref); + full_ref->parent = parent; + full_ref->root = ref_root; + if (parent) + ref->type = BTRFS_SHARED_BLOCK_REF_KEY; + else + ref->type = BTRFS_TREE_BLOCK_REF_KEY; + full_ref->level = level; + + trace_add_delayed_tree_ref(ref, full_ref, action); + + spin_lock(&head_ref->lock); + existing = tree_insert(&head_ref->ref_root, &ref->rb_node); if (existing) { - if (btrfs_delayed_ref_is_head(ref)) - update_existing_head_ref(existing, ref); - else - update_existing_ref(trans, delayed_refs, existing, ref); + update_existing_ref(trans, delayed_refs, head_ref, existing, + ref); + /* + * we've updated the existing ref, free the newly + * allocated ref + */ + kmem_cache_free(btrfs_delayed_tree_ref_cachep, full_ref); + } else { + atomic_inc(&delayed_refs->num_entries); + trans->delayed_ref_updates++; + } + spin_unlock(&head_ref->lock); +} + +/* + * helper to insert a delayed data ref into the rbtree. + */ +static noinline void +add_delayed_data_ref(struct btrfs_fs_info *fs_info, + struct btrfs_trans_handle *trans, + struct btrfs_delayed_ref_head *head_ref, + struct btrfs_delayed_ref_node *ref, u64 bytenr, + u64 num_bytes, u64 parent, u64 ref_root, u64 owner, + u64 offset, int action, int no_quota) +{ + struct btrfs_delayed_ref_node *existing; + struct btrfs_delayed_data_ref *full_ref; + struct btrfs_delayed_ref_root *delayed_refs; + u64 seq = 0; + + if (action == BTRFS_ADD_DELAYED_EXTENT) + action = BTRFS_ADD_DELAYED_REF; + + delayed_refs = &trans->transaction->delayed_refs; + + if (is_fstree(ref_root)) + seq = atomic64_read(&fs_info->tree_mod_seq); + /* first set the basic ref node struct up */ + atomic_set(&ref->refs, 1); + ref->bytenr = bytenr; + ref->num_bytes = num_bytes; + ref->ref_mod = 1; + ref->action = action; + ref->is_head = 0; + ref->in_tree = 1; + ref->no_quota = no_quota; + ref->seq = seq; + + full_ref = btrfs_delayed_node_to_data_ref(ref); + full_ref->parent = parent; + full_ref->root = ref_root; + if (parent) + ref->type = BTRFS_SHARED_DATA_REF_KEY; + else + ref->type = BTRFS_EXTENT_DATA_REF_KEY; + + full_ref->objectid = owner; + full_ref->offset = offset; + + trace_add_delayed_data_ref(ref, full_ref, action); + + spin_lock(&head_ref->lock); + existing = tree_insert(&head_ref->ref_root, &ref->rb_node); + if (existing) { + update_existing_ref(trans, delayed_refs, head_ref, existing, + ref); /* * we've updated the existing ref, free the newly * allocated ref */ - kfree(ref); + kmem_cache_free(btrfs_delayed_data_ref_cachep, full_ref); } else { - if (btrfs_delayed_ref_is_head(ref)) { - delayed_refs->num_heads++; - delayed_refs->num_heads_ready++; - } - delayed_refs->num_entries++; + atomic_inc(&delayed_refs->num_entries); trans->delayed_ref_updates++; } - return 0; + spin_unlock(&head_ref->lock); } /* - * add a delayed ref to the tree. This does all of the accounting required + * add a delayed tree ref. This does all of the accounting required * to make sure the delayed ref is eventually processed before this * transaction commits. */ -int btrfs_add_delayed_ref(struct btrfs_trans_handle *trans, - u64 bytenr, u64 num_bytes, u64 parent, u64 ref_root, - u64 ref_generation, u64 owner_objectid, int action, - int pin) +int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info, + struct btrfs_trans_handle *trans, + u64 bytenr, u64 num_bytes, u64 parent, + u64 ref_root, int level, int action, + struct btrfs_delayed_extent_op *extent_op, + int no_quota) { - struct btrfs_delayed_ref *ref; + struct btrfs_delayed_tree_ref *ref; struct btrfs_delayed_ref_head *head_ref; struct btrfs_delayed_ref_root *delayed_refs; - int ret; - ref = kmalloc(sizeof(*ref), GFP_NOFS); + if (!is_fstree(ref_root) || !fs_info->quota_enabled) + no_quota = 0; + + BUG_ON(extent_op && extent_op->is_data); + ref = kmem_cache_alloc(btrfs_delayed_tree_ref_cachep, GFP_NOFS); if (!ref) return -ENOMEM; - /* - * the parent = 0 case comes from cases where we don't actually - * know the parent yet. It will get updated later via a add/drop - * pair. - */ - if (parent == 0) - parent = bytenr; - - head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS); + head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS); if (!head_ref) { - kfree(ref); + kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref); return -ENOMEM; } + + head_ref->extent_op = extent_op; + delayed_refs = &trans->transaction->delayed_refs; spin_lock(&delayed_refs->lock); @@ -563,84 +796,48 @@ int btrfs_add_delayed_ref(struct btrfs_trans_handle *trans, * insert both the head node and the new ref without dropping * the spin lock */ - ret = __btrfs_add_delayed_ref(trans, &head_ref->node, bytenr, num_bytes, - (u64)-1, 0, 0, 0, action, pin); - BUG_ON(ret); - - ret = __btrfs_add_delayed_ref(trans, &ref->node, bytenr, num_bytes, - parent, ref_root, ref_generation, - owner_objectid, action, pin); - BUG_ON(ret); - spin_unlock(&delayed_refs->lock); - return 0; -} + head_ref = add_delayed_ref_head(fs_info, trans, &head_ref->node, + bytenr, num_bytes, action, 0); -/* - * this does a simple search for the head node for a given extent. - * It must be called with the delayed ref spinlock held, and it returns - * the head node if any where found, or NULL if not. - */ -struct btrfs_delayed_ref_head * -btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr) -{ - struct btrfs_delayed_ref_node *ref; - struct btrfs_delayed_ref_root *delayed_refs; + add_delayed_tree_ref(fs_info, trans, head_ref, &ref->node, bytenr, + num_bytes, parent, ref_root, level, action, + no_quota); + spin_unlock(&delayed_refs->lock); - delayed_refs = &trans->transaction->delayed_refs; - ref = tree_search(&delayed_refs->root, bytenr, (u64)-1, NULL); - if (ref) - return btrfs_delayed_node_to_head(ref); - return NULL; + return 0; } /* - * add a delayed ref to the tree. This does all of the accounting required - * to make sure the delayed ref is eventually processed before this - * transaction commits. - * - * The main point of this call is to add and remove a backreference in a single - * shot, taking the lock only once, and only searching for the head node once. - * - * It is the same as doing a ref add and delete in two separate calls. + * add a delayed data ref. it's similar to btrfs_add_delayed_tree_ref. */ -int btrfs_update_delayed_ref(struct btrfs_trans_handle *trans, - u64 bytenr, u64 num_bytes, u64 orig_parent, - u64 parent, u64 orig_ref_root, u64 ref_root, - u64 orig_ref_generation, u64 ref_generation, - u64 owner_objectid, int pin) +int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info, + struct btrfs_trans_handle *trans, + u64 bytenr, u64 num_bytes, + u64 parent, u64 ref_root, + u64 owner, u64 offset, int action, + struct btrfs_delayed_extent_op *extent_op, + int no_quota) { - struct btrfs_delayed_ref *ref; - struct btrfs_delayed_ref *old_ref; + struct btrfs_delayed_data_ref *ref; struct btrfs_delayed_ref_head *head_ref; struct btrfs_delayed_ref_root *delayed_refs; - int ret; - ref = kmalloc(sizeof(*ref), GFP_NOFS); + if (!is_fstree(ref_root) || !fs_info->quota_enabled) + no_quota = 0; + + BUG_ON(extent_op && !extent_op->is_data); + ref = kmem_cache_alloc(btrfs_delayed_data_ref_cachep, GFP_NOFS); if (!ref) return -ENOMEM; - old_ref = kmalloc(sizeof(*old_ref), GFP_NOFS); - if (!old_ref) { - kfree(ref); + head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS); + if (!head_ref) { + kmem_cache_free(btrfs_delayed_data_ref_cachep, ref); return -ENOMEM; } - /* - * the parent = 0 case comes from cases where we don't actually - * know the parent yet. It will get updated later via a add/drop - * pair. - */ - if (parent == 0) - parent = bytenr; - if (orig_parent == 0) - orig_parent = bytenr; + head_ref->extent_op = extent_op; - head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS); - if (!head_ref) { - kfree(ref); - kfree(old_ref); - return -ENOMEM; - } delayed_refs = &trans->transaction->delayed_refs; spin_lock(&delayed_refs->lock); @@ -648,21 +845,100 @@ int btrfs_update_delayed_ref(struct btrfs_trans_handle *trans, * insert both the head node and the new ref without dropping * the spin lock */ - ret = __btrfs_add_delayed_ref(trans, &head_ref->node, bytenr, num_bytes, - (u64)-1, 0, 0, 0, - BTRFS_UPDATE_DELAYED_HEAD, 0); - BUG_ON(ret); - - ret = __btrfs_add_delayed_ref(trans, &ref->node, bytenr, num_bytes, - parent, ref_root, ref_generation, - owner_objectid, BTRFS_ADD_DELAYED_REF, 0); - BUG_ON(ret); - - ret = __btrfs_add_delayed_ref(trans, &old_ref->node, bytenr, num_bytes, - orig_parent, orig_ref_root, - orig_ref_generation, owner_objectid, - BTRFS_DROP_DELAYED_REF, pin); - BUG_ON(ret); + head_ref = add_delayed_ref_head(fs_info, trans, &head_ref->node, + bytenr, num_bytes, action, 1); + + add_delayed_data_ref(fs_info, trans, head_ref, &ref->node, bytenr, + num_bytes, parent, ref_root, owner, offset, + action, no_quota); spin_unlock(&delayed_refs->lock); + + return 0; +} + +int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info, + struct btrfs_trans_handle *trans, + u64 bytenr, u64 num_bytes, + struct btrfs_delayed_extent_op *extent_op) +{ + struct btrfs_delayed_ref_head *head_ref; + struct btrfs_delayed_ref_root *delayed_refs; + + head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS); + if (!head_ref) + return -ENOMEM; + + head_ref->extent_op = extent_op; + + delayed_refs = &trans->transaction->delayed_refs; + spin_lock(&delayed_refs->lock); + + add_delayed_ref_head(fs_info, trans, &head_ref->node, bytenr, + num_bytes, BTRFS_UPDATE_DELAYED_HEAD, + extent_op->is_data); + + spin_unlock(&delayed_refs->lock); + return 0; +} + +/* + * this does a simple search for the head node for a given extent. + * It must be called with the delayed ref spinlock held, and it returns + * the head node if any where found, or NULL if not. + */ +struct btrfs_delayed_ref_head * +btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr) +{ + struct btrfs_delayed_ref_root *delayed_refs; + + delayed_refs = &trans->transaction->delayed_refs; + return find_ref_head(&delayed_refs->href_root, bytenr, 0); +} + +void btrfs_delayed_ref_exit(void) +{ + if (btrfs_delayed_ref_head_cachep) + kmem_cache_destroy(btrfs_delayed_ref_head_cachep); + if (btrfs_delayed_tree_ref_cachep) + kmem_cache_destroy(btrfs_delayed_tree_ref_cachep); + if (btrfs_delayed_data_ref_cachep) + kmem_cache_destroy(btrfs_delayed_data_ref_cachep); + if (btrfs_delayed_extent_op_cachep) + kmem_cache_destroy(btrfs_delayed_extent_op_cachep); +} + +int btrfs_delayed_ref_init(void) +{ + btrfs_delayed_ref_head_cachep = kmem_cache_create( + "btrfs_delayed_ref_head", + sizeof(struct btrfs_delayed_ref_head), 0, + SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL); + if (!btrfs_delayed_ref_head_cachep) + goto fail; + + btrfs_delayed_tree_ref_cachep = kmem_cache_create( + "btrfs_delayed_tree_ref", + sizeof(struct btrfs_delayed_tree_ref), 0, + SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL); + if (!btrfs_delayed_tree_ref_cachep) + goto fail; + + btrfs_delayed_data_ref_cachep = kmem_cache_create( + "btrfs_delayed_data_ref", + sizeof(struct btrfs_delayed_data_ref), 0, + SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL); + if (!btrfs_delayed_data_ref_cachep) + goto fail; + + btrfs_delayed_extent_op_cachep = kmem_cache_create( + "btrfs_delayed_extent_op", + sizeof(struct btrfs_delayed_extent_op), 0, + SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL); + if (!btrfs_delayed_extent_op_cachep) + goto fail; + return 0; +fail: + btrfs_delayed_ref_exit(); + return -ENOMEM; } |