#include <linux/pagemap.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>
+#include <linux/version.h>
+#include "compat.h"
#include "hash.h"
#include "crc32c.h"
#include "ctree.h"
#include "volumes.h"
#include "locking.h"
#include "ref-cache.h"
+#include "compat.h"
#define PENDING_EXTENT_INSERT 0
#define PENDING_EXTENT_DELETE 1
u64 generation;
u64 orig_generation;
int level;
+ struct list_head list;
+ int del;
};
-static int finish_current_insert(struct btrfs_trans_handle *trans, struct
- btrfs_root *extent_root, int all);
-static int del_pending_extents(struct btrfs_trans_handle *trans, struct
- btrfs_root *extent_root, int all);
-static struct btrfs_block_group_cache *
-__btrfs_find_block_group(struct btrfs_root *root,
- struct btrfs_block_group_cache *hint,
- u64 search_start, int data, int owner);
+static int finish_current_insert(struct btrfs_trans_handle *trans,
+ struct btrfs_root *extent_root, int all);
+static int del_pending_extents(struct btrfs_trans_handle *trans,
+ struct btrfs_root *extent_root, int all);
+static int pin_down_bytes(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ u64 bytenr, u64 num_bytes, int is_data);
+static int update_block_group(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ u64 bytenr, u64 num_bytes, int alloc,
+ int mark_free);
static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
{
* this adds the block group to the fs_info rb tree for the block group
* cache
*/
-int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
+static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
struct btrfs_block_group_cache *block_group)
{
struct rb_node **p;
break;
}
}
+ if (ret)
+ atomic_inc(&ret->count);
spin_unlock(&info->block_group_cache_lock);
return ret;
start = extent_end + 1;
} else if (extent_start > start && extent_start < end) {
size = extent_start - start;
- ret = btrfs_add_free_space_lock(block_group, start,
- size);
+ ret = btrfs_add_free_space(block_group, start,
+ size);
BUG_ON(ret);
start = extent_end + 1;
} else {
if (start < end) {
size = end - start;
- ret = btrfs_add_free_space_lock(block_group, start, size);
+ ret = btrfs_add_free_space(block_group, start, size);
BUG_ON(ret);
}
mutex_unlock(&info->pinned_mutex);
return 0;
}
+static int remove_sb_from_cache(struct btrfs_root *root,
+ struct btrfs_block_group_cache *cache)
+{
+ u64 bytenr;
+ u64 *logical;
+ int stripe_len;
+ int i, nr, ret;
+
+ for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
+ bytenr = btrfs_sb_offset(i);
+ ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
+ cache->key.objectid, bytenr, 0,
+ &logical, &nr, &stripe_len);
+ BUG_ON(ret);
+ while (nr--) {
+ btrfs_remove_free_space(cache, logical[nr],
+ stripe_len);
+ }
+ kfree(logical);
+ }
+ return 0;
+}
+
static int cache_block_group(struct btrfs_root *root,
struct btrfs_block_group_cache *block_group)
{
struct btrfs_key key;
struct extent_buffer *leaf;
int slot;
- u64 last = 0;
- u64 first_free;
- int found = 0;
+ u64 last;
if (!block_group)
return 0;
* skip the locking here
*/
path->skip_locking = 1;
- first_free = max_t(u64, block_group->key.objectid,
- BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE);
- key.objectid = block_group->key.objectid;
+ last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
+ key.objectid = last;
key.offset = 0;
btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
if (ret < 0)
goto err;
- ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
- if (ret < 0)
- goto err;
- if (ret == 0) {
- leaf = path->nodes[0];
- btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
- if (key.objectid + key.offset > first_free)
- first_free = key.objectid + key.offset;
- }
- while(1) {
+
+ while (1) {
leaf = path->nodes[0];
slot = path->slots[0];
if (slot >= btrfs_header_nritems(leaf)) {
break;
if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
- if (!found) {
- last = first_free;
- found = 1;
- }
-
add_new_free_space(block_group, root->fs_info, last,
key.objectid);
path->slots[0]++;
}
- if (!found)
- last = first_free;
-
add_new_free_space(block_group, root->fs_info, last,
block_group->key.objectid +
block_group->key.offset);
+ remove_sb_from_cache(root, block_group);
block_group->cached = 1;
ret = 0;
err:
/*
* return the block group that starts at or after bytenr
*/
-struct btrfs_block_group_cache *btrfs_lookup_first_block_group(struct
- btrfs_fs_info *info,
- u64 bytenr)
+static struct btrfs_block_group_cache *
+btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr)
{
struct btrfs_block_group_cache *cache;
/*
* return the block group that contains teh given bytenr
*/
-struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
- btrfs_fs_info *info,
- u64 bytenr)
+struct btrfs_block_group_cache *btrfs_lookup_block_group(
+ struct btrfs_fs_info *info,
+ u64 bytenr)
{
struct btrfs_block_group_cache *cache;
return cache;
}
+static inline void put_block_group(struct btrfs_block_group_cache *cache)
+{
+ if (atomic_dec_and_test(&cache->count))
+ kfree(cache);
+}
+
static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
u64 flags)
{
return num;
}
-static struct btrfs_block_group_cache *
-__btrfs_find_block_group(struct btrfs_root *root,
- struct btrfs_block_group_cache *hint,
- u64 search_start, int data, int owner)
+u64 btrfs_find_block_group(struct btrfs_root *root,
+ u64 search_start, u64 search_hint, int owner)
{
struct btrfs_block_group_cache *cache;
- struct btrfs_block_group_cache *found_group = NULL;
- struct btrfs_fs_info *info = root->fs_info;
u64 used;
- u64 last = 0;
- u64 free_check;
+ u64 last = max(search_hint, search_start);
+ u64 group_start = 0;
int full_search = 0;
- int factor = 10;
+ int factor = 9;
int wrapped = 0;
-
- if (data & BTRFS_BLOCK_GROUP_METADATA)
- factor = 9;
-
- if (search_start) {
- struct btrfs_block_group_cache *shint;
- shint = btrfs_lookup_first_block_group(info, search_start);
- if (shint && block_group_bits(shint, data) && !shint->ro) {
- spin_lock(&shint->lock);
- used = btrfs_block_group_used(&shint->item);
- if (used + shint->pinned + shint->reserved <
- div_factor(shint->key.offset, factor)) {
- spin_unlock(&shint->lock);
- return shint;
- }
- spin_unlock(&shint->lock);
- }
- }
- if (hint && !hint->ro && block_group_bits(hint, data)) {
- spin_lock(&hint->lock);
- used = btrfs_block_group_used(&hint->item);
- if (used + hint->pinned + hint->reserved <
- div_factor(hint->key.offset, factor)) {
- spin_unlock(&hint->lock);
- return hint;
- }
- spin_unlock(&hint->lock);
- last = hint->key.objectid + hint->key.offset;
- } else {
- if (hint)
- last = max(hint->key.objectid, search_start);
- else
- last = search_start;
- }
again:
while (1) {
cache = btrfs_lookup_first_block_group(root->fs_info, last);
last = cache->key.objectid + cache->key.offset;
used = btrfs_block_group_used(&cache->item);
- if (!cache->ro && block_group_bits(cache, data)) {
- free_check = div_factor(cache->key.offset, factor);
+ if ((full_search || !cache->ro) &&
+ block_group_bits(cache, BTRFS_BLOCK_GROUP_METADATA)) {
if (used + cache->pinned + cache->reserved <
- free_check) {
- found_group = cache;
+ div_factor(cache->key.offset, factor)) {
+ group_start = cache->key.objectid;
spin_unlock(&cache->lock);
+ put_block_group(cache);
goto found;
}
}
spin_unlock(&cache->lock);
+ put_block_group(cache);
cond_resched();
}
if (!wrapped) {
goto again;
}
found:
- return found_group;
-}
-
-struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
- struct btrfs_block_group_cache
- *hint, u64 search_start,
- int data, int owner)
-{
-
- struct btrfs_block_group_cache *ret;
- ret = __btrfs_find_block_group(root, hint, search_start, data, owner);
- return ret;
+ return group_start;
}
/* simple helper to search for an existing extent at a given offset */
* to the key objectid.
*/
-static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans,
+static noinline int lookup_extent_backref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
u64 bytenr, u64 parent,
return ret;
}
-static int noinline insert_extent_backref(struct btrfs_trans_handle *trans,
+/*
+ * updates all the backrefs that are pending on update_list for the
+ * extent_root
+ */
+static noinline int update_backrefs(struct btrfs_trans_handle *trans,
+ struct btrfs_root *extent_root,
+ struct btrfs_path *path,
+ struct list_head *update_list)
+{
+ struct btrfs_key key;
+ struct btrfs_extent_ref *ref;
+ struct btrfs_fs_info *info = extent_root->fs_info;
+ struct pending_extent_op *op;
+ struct extent_buffer *leaf;
+ int ret = 0;
+ struct list_head *cur = update_list->next;
+ u64 ref_objectid;
+ u64 ref_root = extent_root->root_key.objectid;
+
+ op = list_entry(cur, struct pending_extent_op, list);
+
+search:
+ key.objectid = op->bytenr;
+ key.type = BTRFS_EXTENT_REF_KEY;
+ key.offset = op->orig_parent;
+
+ ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 1);
+ BUG_ON(ret);
+
+ leaf = path->nodes[0];
+
+loop:
+ ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref);
+
+ ref_objectid = btrfs_ref_objectid(leaf, ref);
+
+ if (btrfs_ref_root(leaf, ref) != ref_root ||
+ btrfs_ref_generation(leaf, ref) != op->orig_generation ||
+ (ref_objectid != op->level &&
+ ref_objectid != BTRFS_MULTIPLE_OBJECTIDS)) {
+ printk(KERN_ERR "btrfs couldn't find %llu, parent %llu, "
+ "root %llu, owner %u\n",
+ (unsigned long long)op->bytenr,
+ (unsigned long long)op->orig_parent,
+ (unsigned long long)ref_root, op->level);
+ btrfs_print_leaf(extent_root, leaf);
+ BUG();
+ }
+
+ key.objectid = op->bytenr;
+ key.offset = op->parent;
+ key.type = BTRFS_EXTENT_REF_KEY;
+ ret = btrfs_set_item_key_safe(trans, extent_root, path, &key);
+ BUG_ON(ret);
+ ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref);
+ btrfs_set_ref_generation(leaf, ref, op->generation);
+
+ cur = cur->next;
+
+ list_del_init(&op->list);
+ unlock_extent(&info->extent_ins, op->bytenr,
+ op->bytenr + op->num_bytes - 1, GFP_NOFS);
+ kfree(op);
+
+ if (cur == update_list) {
+ btrfs_mark_buffer_dirty(path->nodes[0]);
+ btrfs_release_path(extent_root, path);
+ goto out;
+ }
+
+ op = list_entry(cur, struct pending_extent_op, list);
+
+ path->slots[0]++;
+ while (path->slots[0] < btrfs_header_nritems(leaf)) {
+ btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+ if (key.objectid == op->bytenr &&
+ key.type == BTRFS_EXTENT_REF_KEY)
+ goto loop;
+ path->slots[0]++;
+ }
+
+ btrfs_mark_buffer_dirty(path->nodes[0]);
+ btrfs_release_path(extent_root, path);
+ goto search;
+
+out:
+ return 0;
+}
+
+static noinline int insert_extents(struct btrfs_trans_handle *trans,
+ struct btrfs_root *extent_root,
+ struct btrfs_path *path,
+ struct list_head *insert_list, int nr)
+{
+ struct btrfs_key *keys;
+ u32 *data_size;
+ struct pending_extent_op *op;
+ struct extent_buffer *leaf;
+ struct list_head *cur = insert_list->next;
+ struct btrfs_fs_info *info = extent_root->fs_info;
+ u64 ref_root = extent_root->root_key.objectid;
+ int i = 0, last = 0, ret;
+ int total = nr * 2;
+
+ if (!nr)
+ return 0;
+
+ keys = kzalloc(total * sizeof(struct btrfs_key), GFP_NOFS);
+ if (!keys)
+ return -ENOMEM;
+
+ data_size = kzalloc(total * sizeof(u32), GFP_NOFS);
+ if (!data_size) {
+ kfree(keys);
+ return -ENOMEM;
+ }
+
+ list_for_each_entry(op, insert_list, list) {
+ keys[i].objectid = op->bytenr;
+ keys[i].offset = op->num_bytes;
+ keys[i].type = BTRFS_EXTENT_ITEM_KEY;
+ data_size[i] = sizeof(struct btrfs_extent_item);
+ i++;
+
+ keys[i].objectid = op->bytenr;
+ keys[i].offset = op->parent;
+ keys[i].type = BTRFS_EXTENT_REF_KEY;
+ data_size[i] = sizeof(struct btrfs_extent_ref);
+ i++;
+ }
+
+ op = list_entry(cur, struct pending_extent_op, list);
+ i = 0;
+ while (i < total) {
+ int c;
+ ret = btrfs_insert_some_items(trans, extent_root, path,
+ keys+i, data_size+i, total-i);
+ BUG_ON(ret < 0);
+
+ if (last && ret > 1)
+ BUG();
+
+ leaf = path->nodes[0];
+ for (c = 0; c < ret; c++) {
+ int ref_first = keys[i].type == BTRFS_EXTENT_REF_KEY;
+
+ /*
+ * if the first item we inserted was a backref, then
+ * the EXTENT_ITEM will be the odd c's, else it will
+ * be the even c's
+ */
+ if ((ref_first && (c % 2)) ||
+ (!ref_first && !(c % 2))) {
+ struct btrfs_extent_item *itm;
+
+ itm = btrfs_item_ptr(leaf, path->slots[0] + c,
+ struct btrfs_extent_item);
+ btrfs_set_extent_refs(path->nodes[0], itm, 1);
+ op->del++;
+ } else {
+ struct btrfs_extent_ref *ref;
+
+ ref = btrfs_item_ptr(leaf, path->slots[0] + c,
+ struct btrfs_extent_ref);
+ btrfs_set_ref_root(leaf, ref, ref_root);
+ btrfs_set_ref_generation(leaf, ref,
+ op->generation);
+ btrfs_set_ref_objectid(leaf, ref, op->level);
+ btrfs_set_ref_num_refs(leaf, ref, 1);
+ op->del++;
+ }
+
+ /*
+ * using del to see when its ok to free up the
+ * pending_extent_op. In the case where we insert the
+ * last item on the list in order to help do batching
+ * we need to not free the extent op until we actually
+ * insert the extent_item
+ */
+ if (op->del == 2) {
+ unlock_extent(&info->extent_ins, op->bytenr,
+ op->bytenr + op->num_bytes - 1,
+ GFP_NOFS);
+ cur = cur->next;
+ list_del_init(&op->list);
+ kfree(op);
+ if (cur != insert_list)
+ op = list_entry(cur,
+ struct pending_extent_op,
+ list);
+ }
+ }
+ btrfs_mark_buffer_dirty(leaf);
+ btrfs_release_path(extent_root, path);
+
+ /*
+ * Ok backref's and items usually go right next to eachother,
+ * but if we could only insert 1 item that means that we
+ * inserted on the end of a leaf, and we have no idea what may
+ * be on the next leaf so we just play it safe. In order to
+ * try and help this case we insert the last thing on our
+ * insert list so hopefully it will end up being the last
+ * thing on the leaf and everything else will be before it,
+ * which will let us insert a whole bunch of items at the same
+ * time.
+ */
+ if (ret == 1 && !last && (i + ret < total)) {
+ /*
+ * last: where we will pick up the next time around
+ * i: our current key to insert, will be total - 1
+ * cur: the current op we are screwing with
+ * op: duh
+ */
+ last = i + ret;
+ i = total - 1;
+ cur = insert_list->prev;
+ op = list_entry(cur, struct pending_extent_op, list);
+ } else if (last) {
+ /*
+ * ok we successfully inserted the last item on the
+ * list, lets reset everything
+ *
+ * i: our current key to insert, so where we left off
+ * last time
+ * last: done with this
+ * cur: the op we are messing with
+ * op: duh
+ * total: since we inserted the last key, we need to
+ * decrement total so we dont overflow
+ */
+ i = last;
+ last = 0;
+ total--;
+ if (i < total) {
+ cur = insert_list->next;
+ op = list_entry(cur, struct pending_extent_op,
+ list);
+ }
+ } else {
+ i += ret;
+ }
+
+ cond_resched();
+ }
+ ret = 0;
+ kfree(keys);
+ kfree(data_size);
+ return ret;
+}
+
+static noinline int insert_extent_backref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
u64 bytenr, u64 parent,
return ret;
}
-static int noinline remove_extent_backref(struct btrfs_trans_handle *trans,
+static noinline int remove_extent_backref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path)
{
return ret;
}
+#ifdef BIO_RW_DISCARD
+static void btrfs_issue_discard(struct block_device *bdev,
+ u64 start, u64 len)
+{
+ blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL);
+}
+#endif
+
+static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
+ u64 num_bytes)
+{
+#ifdef BIO_RW_DISCARD
+ int ret;
+ u64 map_length = num_bytes;
+ struct btrfs_multi_bio *multi = NULL;
+
+ /* Tell the block device(s) that the sectors can be discarded */
+ ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
+ bytenr, &map_length, &multi, 0);
+ if (!ret) {
+ struct btrfs_bio_stripe *stripe = multi->stripes;
+ int i;
+
+ if (map_length > num_bytes)
+ map_length = num_bytes;
+
+ for (i = 0; i < multi->num_stripes; i++, stripe++) {
+ btrfs_issue_discard(stripe->dev->bdev,
+ stripe->physical,
+ map_length);
+ }
+ kfree(multi);
+ }
+
+ return ret;
+#else
+ return 0;
+#endif
+}
+
+static noinline int free_extents(struct btrfs_trans_handle *trans,
+ struct btrfs_root *extent_root,
+ struct list_head *del_list)
+{
+ struct btrfs_fs_info *info = extent_root->fs_info;
+ struct btrfs_path *path;
+ struct btrfs_key key, found_key;
+ struct extent_buffer *leaf;
+ struct list_head *cur;
+ struct pending_extent_op *op;
+ struct btrfs_extent_item *ei;
+ int ret, num_to_del, extent_slot = 0, found_extent = 0;
+ u32 refs;
+ u64 bytes_freed = 0;
+
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
+ path->reada = 1;
+
+search:
+ /* search for the backref for the current ref we want to delete */
+ cur = del_list->next;
+ op = list_entry(cur, struct pending_extent_op, list);
+ ret = lookup_extent_backref(trans, extent_root, path, op->bytenr,
+ op->orig_parent,
+ extent_root->root_key.objectid,
+ op->orig_generation, op->level, 1);
+ if (ret) {
+ printk(KERN_ERR "btrfs unable to find backref byte nr %llu "
+ "root %llu gen %llu owner %u\n",
+ (unsigned long long)op->bytenr,
+ (unsigned long long)extent_root->root_key.objectid,
+ (unsigned long long)op->orig_generation, op->level);
+ btrfs_print_leaf(extent_root, path->nodes[0]);
+ WARN_ON(1);
+ goto out;
+ }
+
+ extent_slot = path->slots[0];
+ num_to_del = 1;
+ found_extent = 0;
+
+ /*
+ * if we aren't the first item on the leaf we can move back one and see
+ * if our ref is right next to our extent item
+ */
+ if (likely(extent_slot)) {
+ extent_slot--;
+ btrfs_item_key_to_cpu(path->nodes[0], &found_key,
+ extent_slot);
+ if (found_key.objectid == op->bytenr &&
+ found_key.type == BTRFS_EXTENT_ITEM_KEY &&
+ found_key.offset == op->num_bytes) {
+ num_to_del++;
+ found_extent = 1;
+ }
+ }
+
+ /*
+ * if we didn't find the extent we need to delete the backref and then
+ * search for the extent item key so we can update its ref count
+ */
+ if (!found_extent) {
+ key.objectid = op->bytenr;
+ key.type = BTRFS_EXTENT_ITEM_KEY;
+ key.offset = op->num_bytes;
+
+ ret = remove_extent_backref(trans, extent_root, path);
+ BUG_ON(ret);
+ btrfs_release_path(extent_root, path);
+ ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1);
+ BUG_ON(ret);
+ extent_slot = path->slots[0];
+ }
+
+ /* this is where we update the ref count for the extent */
+ leaf = path->nodes[0];
+ ei = btrfs_item_ptr(leaf, extent_slot, struct btrfs_extent_item);
+ refs = btrfs_extent_refs(leaf, ei);
+ BUG_ON(refs == 0);
+ refs--;
+ btrfs_set_extent_refs(leaf, ei, refs);
+
+ btrfs_mark_buffer_dirty(leaf);
+
+ /*
+ * This extent needs deleting. The reason cur_slot is extent_slot +
+ * num_to_del is because extent_slot points to the slot where the extent
+ * is, and if the backref was not right next to the extent we will be
+ * deleting at least 1 item, and will want to start searching at the
+ * slot directly next to extent_slot. However if we did find the
+ * backref next to the extent item them we will be deleting at least 2
+ * items and will want to start searching directly after the ref slot
+ */
+ if (!refs) {
+ struct list_head *pos, *n, *end;
+ int cur_slot = extent_slot+num_to_del;
+ u64 super_used;
+ u64 root_used;
+
+ path->slots[0] = extent_slot;
+ bytes_freed = op->num_bytes;
+
+ mutex_lock(&info->pinned_mutex);
+ ret = pin_down_bytes(trans, extent_root, op->bytenr,
+ op->num_bytes, op->level >=
+ BTRFS_FIRST_FREE_OBJECTID);
+ mutex_unlock(&info->pinned_mutex);
+ BUG_ON(ret < 0);
+ op->del = ret;
+
+ /*
+ * we need to see if we can delete multiple things at once, so
+ * start looping through the list of extents we are wanting to
+ * delete and see if their extent/backref's are right next to
+ * eachother and the extents only have 1 ref
+ */
+ for (pos = cur->next; pos != del_list; pos = pos->next) {
+ struct pending_extent_op *tmp;
+
+ tmp = list_entry(pos, struct pending_extent_op, list);
+
+ /* we only want to delete extent+ref at this stage */
+ if (cur_slot >= btrfs_header_nritems(leaf) - 1)
+ break;
+
+ btrfs_item_key_to_cpu(leaf, &found_key, cur_slot);
+ if (found_key.objectid != tmp->bytenr ||
+ found_key.type != BTRFS_EXTENT_ITEM_KEY ||
+ found_key.offset != tmp->num_bytes)
+ break;
+
+ /* check to make sure this extent only has one ref */
+ ei = btrfs_item_ptr(leaf, cur_slot,
+ struct btrfs_extent_item);
+ if (btrfs_extent_refs(leaf, ei) != 1)
+ break;
+
+ btrfs_item_key_to_cpu(leaf, &found_key, cur_slot+1);
+ if (found_key.objectid != tmp->bytenr ||
+ found_key.type != BTRFS_EXTENT_REF_KEY ||
+ found_key.offset != tmp->orig_parent)
+ break;
+
+ /*
+ * the ref is right next to the extent, we can set the
+ * ref count to 0 since we will delete them both now
+ */
+ btrfs_set_extent_refs(leaf, ei, 0);
+
+ /* pin down the bytes for this extent */
+ mutex_lock(&info->pinned_mutex);
+ ret = pin_down_bytes(trans, extent_root, tmp->bytenr,
+ tmp->num_bytes, tmp->level >=
+ BTRFS_FIRST_FREE_OBJECTID);
+ mutex_unlock(&info->pinned_mutex);
+ BUG_ON(ret < 0);
+
+ /*
+ * use the del field to tell if we need to go ahead and
+ * free up the extent when we delete the item or not.
+ */
+ tmp->del = ret;
+ bytes_freed += tmp->num_bytes;
+
+ num_to_del += 2;
+ cur_slot += 2;
+ }
+ end = pos;
+
+ /* update the free space counters */
+ spin_lock(&info->delalloc_lock);
+ super_used = btrfs_super_bytes_used(&info->super_copy);
+ btrfs_set_super_bytes_used(&info->super_copy,
+ super_used - bytes_freed);
+
+ root_used = btrfs_root_used(&extent_root->root_item);
+ btrfs_set_root_used(&extent_root->root_item,
+ root_used - bytes_freed);
+ spin_unlock(&info->delalloc_lock);
+
+ /* delete the items */
+ ret = btrfs_del_items(trans, extent_root, path,
+ path->slots[0], num_to_del);
+ BUG_ON(ret);
+
+ /*
+ * loop through the extents we deleted and do the cleanup work
+ * on them
+ */
+ for (pos = cur, n = pos->next; pos != end;
+ pos = n, n = pos->next) {
+ struct pending_extent_op *tmp;
+ tmp = list_entry(pos, struct pending_extent_op, list);
+
+ /*
+ * remember tmp->del tells us wether or not we pinned
+ * down the extent
+ */
+ ret = update_block_group(trans, extent_root,
+ tmp->bytenr, tmp->num_bytes, 0,
+ tmp->del);
+ BUG_ON(ret);
+
+ list_del_init(&tmp->list);
+ unlock_extent(&info->extent_ins, tmp->bytenr,
+ tmp->bytenr + tmp->num_bytes - 1,
+ GFP_NOFS);
+ kfree(tmp);
+ }
+ } else if (refs && found_extent) {
+ /*
+ * the ref and extent were right next to eachother, but the
+ * extent still has a ref, so just free the backref and keep
+ * going
+ */
+ ret = remove_extent_backref(trans, extent_root, path);
+ BUG_ON(ret);
+
+ list_del_init(&op->list);
+ unlock_extent(&info->extent_ins, op->bytenr,
+ op->bytenr + op->num_bytes - 1, GFP_NOFS);
+ kfree(op);
+ } else {
+ /*
+ * the extent has multiple refs and the backref we were looking
+ * for was not right next to it, so just unlock and go next,
+ * we're good to go
+ */
+ list_del_init(&op->list);
+ unlock_extent(&info->extent_ins, op->bytenr,
+ op->bytenr + op->num_bytes - 1, GFP_NOFS);
+ kfree(op);
+ }
+
+ btrfs_release_path(extent_root, path);
+ if (!list_empty(del_list))
+ goto search;
+
+out:
+ btrfs_free_path(path);
+ return ret;
+}
+
static int __btrfs_update_extent_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 bytenr,
u64 orig_parent, u64 parent,
extent_op->generation = ref_generation;
extent_op->orig_generation = orig_generation;
extent_op->level = (int)owner_objectid;
+ INIT_LIST_HEAD(&extent_op->list);
+ extent_op->del = 0;
set_extent_bits(&root->fs_info->extent_ins,
bytenr, bytenr + num_bytes - 1,
btrfs_item_key_to_cpu(l, &key, path->slots[0]);
if (key.objectid != bytenr) {
btrfs_print_leaf(root->fs_info->extent_root, path->nodes[0]);
- printk("wanted %Lu found %Lu\n", bytenr, key.objectid);
+ printk(KERN_ERR "btrfs wanted %llu found %llu\n",
+ (unsigned long long)bytenr,
+ (unsigned long long)key.objectid);
BUG();
}
BUG_ON(key.type != BTRFS_EXTENT_ITEM_KEY);
goto out;
if (ret != 0) {
btrfs_print_leaf(root, path->nodes[0]);
- printk("failed to find block number %Lu\n", bytenr);
+ printk(KERN_INFO "btrfs failed to find block number %llu\n",
+ (unsigned long long)bytenr);
BUG();
}
l = path->nodes[0];
}
int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
- struct btrfs_root *root, u64 bytenr)
+ struct btrfs_root *root, u64 objectid, u64 bytenr)
{
struct btrfs_root *extent_root = root->fs_info->extent_root;
struct btrfs_path *path;
ref_item = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_ref);
ref_root = btrfs_ref_root(leaf, ref_item);
- if (ref_root != root->root_key.objectid &&
- ref_root != BTRFS_TREE_LOG_OBJECTID) {
+ if ((ref_root != root->root_key.objectid &&
+ ref_root != BTRFS_TREE_LOG_OBJECTID) ||
+ objectid != btrfs_ref_objectid(leaf, ref_item)) {
ret = 1;
goto out;
}
if (!path)
return -ENOMEM;
- while(1) {
+ while (1) {
cache = NULL;
spin_lock(&root->fs_info->block_group_cache_lock);
for (n = rb_first(&root->fs_info->block_group_cache_tree);
return werr;
}
+int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
+{
+ struct btrfs_block_group_cache *block_group;
+ int readonly = 0;
+
+ block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
+ if (!block_group || block_group->ro)
+ readonly = 1;
+ if (block_group)
+ put_block_group(block_group);
+ return readonly;
+}
+
static int update_space_info(struct btrfs_fs_info *info, u64 flags,
u64 total_bytes, u64 bytes_used,
struct btrfs_space_info **space_info)
*space_info = found;
return 0;
}
- found = kmalloc(sizeof(*found), GFP_NOFS);
+ found = kzalloc(sizeof(*found), GFP_NOFS);
if (!found)
return -ENOMEM;
found->bytes_used = bytes_used;
found->bytes_pinned = 0;
found->bytes_reserved = 0;
+ found->bytes_readonly = 0;
found->full = 0;
found->force_alloc = 0;
*space_info = found;
}
}
-static u64 reduce_alloc_profile(struct btrfs_root *root, u64 flags)
+static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
+{
+ spin_lock(&cache->space_info->lock);
+ spin_lock(&cache->lock);
+ if (!cache->ro) {
+ cache->space_info->bytes_readonly += cache->key.offset -
+ btrfs_block_group_used(&cache->item);
+ cache->ro = 1;
+ }
+ spin_unlock(&cache->lock);
+ spin_unlock(&cache->space_info->lock);
+}
+
+u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
{
- u64 num_devices = root->fs_info->fs_devices->num_devices;
+ u64 num_devices = root->fs_info->fs_devices->rw_devices;
if (num_devices == 1)
flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
{
struct btrfs_space_info *space_info;
u64 thresh;
- u64 start;
- u64 num_bytes;
- int ret = 0, waited = 0;
+ int ret = 0;
+
+ mutex_lock(&extent_root->fs_info->chunk_mutex);
- flags = reduce_alloc_profile(extent_root, flags);
+ flags = btrfs_reduce_alloc_profile(extent_root, flags);
space_info = __find_space_info(extent_root->fs_info, flags);
if (!space_info) {
goto out;
}
- thresh = div_factor(space_info->total_bytes, 6);
+ thresh = space_info->total_bytes - space_info->bytes_readonly;
+ thresh = div_factor(thresh, 6);
if (!force &&
(space_info->bytes_used + space_info->bytes_pinned +
space_info->bytes_reserved + alloc_bytes) < thresh) {
spin_unlock(&space_info->lock);
goto out;
}
-
spin_unlock(&space_info->lock);
- ret = mutex_trylock(&extent_root->fs_info->chunk_mutex);
- if (!ret && !force) {
- goto out;
- } else if (!ret) {
- mutex_lock(&extent_root->fs_info->chunk_mutex);
- waited = 1;
- }
-
- if (waited) {
- spin_lock(&space_info->lock);
- if (space_info->full) {
- spin_unlock(&space_info->lock);
- goto out_unlock;
- }
- spin_unlock(&space_info->lock);
- }
-
- ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags);
- if (ret) {
-printk("space info full %Lu\n", flags);
+ ret = btrfs_alloc_chunk(trans, extent_root, flags);
+ if (ret)
space_info->full = 1;
- goto out_unlock;
- }
-
- ret = btrfs_make_block_group(trans, extent_root, 0, flags,
- BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, num_bytes);
- BUG_ON(ret);
-out_unlock:
- mutex_unlock(&extent_root->fs_info->chunk_mutex);
out:
+ mutex_unlock(&extent_root->fs_info->chunk_mutex);
return ret;
}
u64 old_val;
u64 byte_in_group;
- while(total) {
+ while (total) {
cache = btrfs_lookup_block_group(info, bytenr);
- if (!cache) {
+ if (!cache)
return -1;
- }
byte_in_group = bytenr - cache->key.objectid;
WARN_ON(byte_in_group > cache->key.offset);
if (alloc) {
old_val += num_bytes;
cache->space_info->bytes_used += num_bytes;
+ if (cache->ro)
+ cache->space_info->bytes_readonly -= num_bytes;
btrfs_set_block_group_used(&cache->item, old_val);
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
} else {
old_val -= num_bytes;
cache->space_info->bytes_used -= num_bytes;
+ if (cache->ro)
+ cache->space_info->bytes_readonly += num_bytes;
btrfs_set_block_group_used(&cache->item, old_val);
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
if (mark_free) {
int ret;
+
+ ret = btrfs_discard_extent(root, bytenr,
+ num_bytes);
+ WARN_ON(ret);
+
ret = btrfs_add_free_space(cache, bytenr,
num_bytes);
- if (ret)
- return -1;
+ WARN_ON(ret);
}
}
+ put_block_group(cache);
total -= num_bytes;
bytenr += num_bytes;
}
static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
{
struct btrfs_block_group_cache *cache;
+ u64 bytenr;
cache = btrfs_lookup_first_block_group(root->fs_info, search_start);
if (!cache)
return 0;
- return cache->key.objectid;
+ bytenr = cache->key.objectid;
+ put_block_group(cache);
+
+ return bytenr;
}
int btrfs_update_pinned_extents(struct btrfs_root *root,
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
fs_info->total_pinned -= len;
+ if (cache->cached)
+ btrfs_add_free_space(cache, bytenr, len);
}
+ put_block_group(cache);
bytenr += len;
num -= len;
}
}
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
+ put_block_group(cache);
bytenr += len;
num -= len;
}
int ret;
mutex_lock(&root->fs_info->pinned_mutex);
- while(1) {
+ while (1) {
ret = find_first_extent_bit(pinned_extents, last,
&start, &end, EXTENT_DIRTY);
if (ret)
u64 start;
u64 end;
int ret;
- struct btrfs_block_group_cache *cache;
mutex_lock(&root->fs_info->pinned_mutex);
- while(1) {
+ while (1) {
ret = find_first_extent_bit(unpin, 0, &start, &end,
EXTENT_DIRTY);
if (ret)
break;
+
+ ret = btrfs_discard_extent(root, start, end + 1 - start);
+
btrfs_update_pinned_extents(root, start, end + 1 - start, 0);
clear_extent_dirty(unpin, start, end, GFP_NOFS);
- cache = btrfs_lookup_block_group(root->fs_info, start);
- if (cache->cached)
- btrfs_add_free_space(cache, start, end - start + 1);
+
if (need_resched()) {
mutex_unlock(&root->fs_info->pinned_mutex);
cond_resched();
}
}
mutex_unlock(&root->fs_info->pinned_mutex);
- return 0;
+ return ret;
}
static int finish_current_insert(struct btrfs_trans_handle *trans,
u64 end;
u64 priv;
u64 search = 0;
+ u64 skipped = 0;
struct btrfs_fs_info *info = extent_root->fs_info;
struct btrfs_path *path;
- struct btrfs_extent_ref *ref;
- struct pending_extent_op *extent_op;
- struct btrfs_key key;
- struct btrfs_extent_item extent_item;
+ struct pending_extent_op *extent_op, *tmp;
+ struct list_head insert_list, update_list;
int ret;
- int err = 0;
+ int num_inserts = 0, max_inserts;
- btrfs_set_stack_extent_refs(&extent_item, 1);
path = btrfs_alloc_path();
+ INIT_LIST_HEAD(&insert_list);
+ INIT_LIST_HEAD(&update_list);
- while(1) {
- mutex_lock(&info->extent_ins_mutex);
+ max_inserts = extent_root->leafsize /
+ (2 * sizeof(struct btrfs_key) + 2 * sizeof(struct btrfs_item) +
+ sizeof(struct btrfs_extent_ref) +
+ sizeof(struct btrfs_extent_item));
+again:
+ mutex_lock(&info->extent_ins_mutex);
+ while (1) {
ret = find_first_extent_bit(&info->extent_ins, search, &start,
&end, EXTENT_WRITEBACK);
if (ret) {
- mutex_unlock(&info->extent_ins_mutex);
- if (search && all) {
+ if (skipped && all && !num_inserts) {
+ skipped = 0;
search = 0;
continue;
}
+ mutex_unlock(&info->extent_ins_mutex);
break;
}
ret = try_lock_extent(&info->extent_ins, start, end, GFP_NOFS);
if (!ret) {
+ skipped = 1;
search = end + 1;
- mutex_unlock(&info->extent_ins_mutex);
- cond_resched();
+ if (need_resched()) {
+ mutex_unlock(&info->extent_ins_mutex);
+ cond_resched();
+ mutex_lock(&info->extent_ins_mutex);
+ }
continue;
}
- BUG_ON(ret < 0);
ret = get_state_private(&info->extent_ins, start, &priv);
BUG_ON(ret);
- extent_op = (struct pending_extent_op *)(unsigned long)priv;
-
- mutex_unlock(&info->extent_ins_mutex);
+ extent_op = (struct pending_extent_op *)(unsigned long) priv;
if (extent_op->type == PENDING_EXTENT_INSERT) {
- key.objectid = start;
- key.offset = end + 1 - start;
- key.type = BTRFS_EXTENT_ITEM_KEY;
- err = btrfs_insert_item(trans, extent_root, &key,
- &extent_item, sizeof(extent_item));
- BUG_ON(err);
+ num_inserts++;
+ list_add_tail(&extent_op->list, &insert_list);
+ search = end + 1;
+ if (num_inserts == max_inserts) {
+ mutex_unlock(&info->extent_ins_mutex);
+ break;
+ }
+ } else if (extent_op->type == PENDING_BACKREF_UPDATE) {
+ list_add_tail(&extent_op->list, &update_list);
+ search = end + 1;
+ } else {
+ BUG();
+ }
+ }
- mutex_lock(&info->extent_ins_mutex);
- clear_extent_bits(&info->extent_ins, start, end,
- EXTENT_WRITEBACK, GFP_NOFS);
- mutex_unlock(&info->extent_ins_mutex);
+ /*
+ * process the update list, clear the writeback bit for it, and if
+ * somebody marked this thing for deletion then just unlock it and be
+ * done, the free_extents will handle it
+ */
+ mutex_lock(&info->extent_ins_mutex);
+ list_for_each_entry_safe(extent_op, tmp, &update_list, list) {
+ clear_extent_bits(&info->extent_ins, extent_op->bytenr,
+ extent_op->bytenr + extent_op->num_bytes - 1,
+ EXTENT_WRITEBACK, GFP_NOFS);
+ if (extent_op->del) {
+ list_del_init(&extent_op->list);
+ unlock_extent(&info->extent_ins, extent_op->bytenr,
+ extent_op->bytenr + extent_op->num_bytes
+ - 1, GFP_NOFS);
+ kfree(extent_op);
+ }
+ }
+ mutex_unlock(&info->extent_ins_mutex);
- err = insert_extent_backref(trans, extent_root, path,
- start, extent_op->parent,
- extent_root->root_key.objectid,
- extent_op->generation,
- extent_op->level);
- BUG_ON(err);
- } else if (extent_op->type == PENDING_BACKREF_UPDATE) {
- err = lookup_extent_backref(trans, extent_root, path,
- start, extent_op->orig_parent,
- extent_root->root_key.objectid,
- extent_op->orig_generation,
- extent_op->level, 0);
- BUG_ON(err);
+ /*
+ * still have things left on the update list, go ahead an update
+ * everything
+ */
+ if (!list_empty(&update_list)) {
+ ret = update_backrefs(trans, extent_root, path, &update_list);
+ BUG_ON(ret);
+ }
- mutex_lock(&info->extent_ins_mutex);
- clear_extent_bits(&info->extent_ins, start, end,
- EXTENT_WRITEBACK, GFP_NOFS);
- mutex_unlock(&info->extent_ins_mutex);
+ /*
+ * if no inserts need to be done, but we skipped some extents and we
+ * need to make sure everything is cleaned then reset everything and
+ * go back to the beginning
+ */
+ if (!num_inserts && all && skipped) {
+ search = 0;
+ skipped = 0;
+ INIT_LIST_HEAD(&update_list);
+ INIT_LIST_HEAD(&insert_list);
+ goto again;
+ } else if (!num_inserts) {
+ goto out;
+ }
- key.objectid = start;
- key.offset = extent_op->parent;
- key.type = BTRFS_EXTENT_REF_KEY;
- err = btrfs_set_item_key_safe(trans, extent_root, path,
- &key);
- BUG_ON(err);
- ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
- struct btrfs_extent_ref);
- btrfs_set_ref_generation(path->nodes[0], ref,
- extent_op->generation);
- btrfs_mark_buffer_dirty(path->nodes[0]);
- btrfs_release_path(extent_root, path);
- } else {
- BUG_ON(1);
+ /*
+ * process the insert extents list. Again if we are deleting this
+ * extent, then just unlock it, pin down the bytes if need be, and be
+ * done with it. Saves us from having to actually insert the extent
+ * into the tree and then subsequently come along and delete it
+ */
+ mutex_lock(&info->extent_ins_mutex);
+ list_for_each_entry_safe(extent_op, tmp, &insert_list, list) {
+ clear_extent_bits(&info->extent_ins, extent_op->bytenr,
+ extent_op->bytenr + extent_op->num_bytes - 1,
+ EXTENT_WRITEBACK, GFP_NOFS);
+ if (extent_op->del) {
+ u64 used;
+ list_del_init(&extent_op->list);
+ unlock_extent(&info->extent_ins, extent_op->bytenr,
+ extent_op->bytenr + extent_op->num_bytes
+ - 1, GFP_NOFS);
+
+ mutex_lock(&extent_root->fs_info->pinned_mutex);
+ ret = pin_down_bytes(trans, extent_root,
+ extent_op->bytenr,
+ extent_op->num_bytes, 0);
+ mutex_unlock(&extent_root->fs_info->pinned_mutex);
+
+ spin_lock(&info->delalloc_lock);
+ used = btrfs_super_bytes_used(&info->super_copy);
+ btrfs_set_super_bytes_used(&info->super_copy,
+ used - extent_op->num_bytes);
+ used = btrfs_root_used(&extent_root->root_item);
+ btrfs_set_root_used(&extent_root->root_item,
+ used - extent_op->num_bytes);
+ spin_unlock(&info->delalloc_lock);
+
+ ret = update_block_group(trans, extent_root,
+ extent_op->bytenr,
+ extent_op->num_bytes,
+ 0, ret > 0);
+ BUG_ON(ret);
+ kfree(extent_op);
+ num_inserts--;
}
- kfree(extent_op);
- unlock_extent(&info->extent_ins, start, end, GFP_NOFS);
- if (all)
- search = 0;
- else
- search = end + 1;
+ }
+ mutex_unlock(&info->extent_ins_mutex);
- cond_resched();
+ ret = insert_extents(trans, extent_root, path, &insert_list,
+ num_inserts);
+ BUG_ON(ret);
+
+ /*
+ * if we broke out of the loop in order to insert stuff because we hit
+ * the maximum number of inserts at a time we can handle, then loop
+ * back and pick up where we left off
+ */
+ if (num_inserts == max_inserts) {
+ INIT_LIST_HEAD(&insert_list);
+ INIT_LIST_HEAD(&update_list);
+ num_inserts = 0;
+ goto again;
+ }
+
+ /*
+ * again, if we need to make absolutely sure there are no more pending
+ * extent operations left and we know that we skipped some, go back to
+ * the beginning and do it all again
+ */
+ if (all && skipped) {
+ INIT_LIST_HEAD(&insert_list);
+ INIT_LIST_HEAD(&update_list);
+ search = 0;
+ skipped = 0;
+ num_inserts = 0;
+ goto again;
}
+out:
btrfs_free_path(path);
return 0;
}
if (ret == 0) {
struct btrfs_key found_key;
extent_slot = path->slots[0];
- while(extent_slot > 0) {
+ while (extent_slot > 0) {
extent_slot--;
btrfs_item_key_to_cpu(path->nodes[0], &found_key,
extent_slot);
btrfs_release_path(extent_root, path);
ret = btrfs_search_slot(trans, extent_root,
&key, path, -1, 1);
+ if (ret) {
+ printk(KERN_ERR "umm, got %d back from search"
+ ", was looking for %llu\n", ret,
+ (unsigned long long)bytenr);
+ btrfs_print_leaf(extent_root, path->nodes[0]);
+ }
BUG_ON(ret);
extent_slot = path->slots[0];
}
} else {
btrfs_print_leaf(extent_root, path->nodes[0]);
WARN_ON(1);
- printk("Unable to find ref byte nr %Lu root %Lu "
- "gen %Lu owner %Lu\n", bytenr,
- root_objectid, ref_generation, owner_objectid);
+ printk(KERN_ERR "btrfs unable to find ref byte nr %llu "
+ "root %llu gen %llu owner %llu\n",
+ (unsigned long long)bytenr,
+ (unsigned long long)root_objectid,
+ (unsigned long long)ref_generation,
+ (unsigned long long)owner_objectid);
}
leaf = path->nodes[0];
if (refs == 0) {
u64 super_used;
u64 root_used;
-#ifdef BIO_RW_DISCARD
- u64 map_length = num_bytes;
- struct btrfs_multi_bio *multi = NULL;
-#endif
if (pin) {
mutex_lock(&root->fs_info->pinned_mutex);
mark_free = 1;
BUG_ON(ret < 0);
}
-
/* block accounting for super block */
- spin_lock_irq(&info->delalloc_lock);
+ spin_lock(&info->delalloc_lock);
super_used = btrfs_super_bytes_used(&info->super_copy);
btrfs_set_super_bytes_used(&info->super_copy,
super_used - num_bytes);
- spin_unlock_irq(&info->delalloc_lock);
/* block accounting for root item */
root_used = btrfs_root_used(&root->root_item);
btrfs_set_root_used(&root->root_item,
root_used - num_bytes);
+ spin_unlock(&info->delalloc_lock);
ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
num_to_del);
BUG_ON(ret);
btrfs_release_path(extent_root, path);
+
+ if (owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
+ ret = btrfs_del_csums(trans, root, bytenr, num_bytes);
+ BUG_ON(ret);
+ }
+
ret = update_block_group(trans, root, bytenr, num_bytes, 0,
mark_free);
BUG_ON(ret);
-
-#ifdef BIO_RW_DISCARD
- /* Tell the block device(s) that the sectors can be discarded */
- ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
- bytenr, &map_length, &multi, 0);
- if (!ret) {
- struct btrfs_bio_stripe *stripe = multi->stripes;
- int i;
-
- if (map_length > num_bytes)
- map_length = num_bytes;
-
- for (i = 0; i < multi->num_stripes; i++, stripe++) {
- blkdev_issue_discard(stripe->dev->bdev,
- stripe->physical >> 9,
- map_length >> 9);
- }
- kfree(multi);
- }
-#endif
}
btrfs_free_path(path);
finish_current_insert(trans, extent_root, 0);
* find all the blocks marked as pending in the radix tree and remove
* them from the extent map
*/
-static int del_pending_extents(struct btrfs_trans_handle *trans, struct
- btrfs_root *extent_root, int all)
+static int del_pending_extents(struct btrfs_trans_handle *trans,
+ struct btrfs_root *extent_root, int all)
{
int ret;
int err = 0;
u64 end;
u64 priv;
u64 search = 0;
+ int nr = 0, skipped = 0;
struct extent_io_tree *pending_del;
struct extent_io_tree *extent_ins;
struct pending_extent_op *extent_op;
struct btrfs_fs_info *info = extent_root->fs_info;
+ struct list_head delete_list;
+ INIT_LIST_HEAD(&delete_list);
extent_ins = &extent_root->fs_info->extent_ins;
pending_del = &extent_root->fs_info->pending_del;
- while(1) {
- mutex_lock(&info->extent_ins_mutex);
+again:
+ mutex_lock(&info->extent_ins_mutex);
+ while (1) {
ret = find_first_extent_bit(pending_del, search, &start, &end,
EXTENT_WRITEBACK);
if (ret) {
- mutex_unlock(&info->extent_ins_mutex);
- if (all && search) {
+ if (all && skipped && !nr) {
search = 0;
continue;
}
+ mutex_unlock(&info->extent_ins_mutex);
break;
}
ret = try_lock_extent(extent_ins, start, end, GFP_NOFS);
if (!ret) {
search = end+1;
- mutex_unlock(&info->extent_ins_mutex);
- cond_resched();
+ skipped = 1;
+
+ if (need_resched()) {
+ mutex_unlock(&info->extent_ins_mutex);
+ cond_resched();
+ mutex_lock(&info->extent_ins_mutex);
+ }
+
continue;
}
BUG_ON(ret < 0);
GFP_NOFS);
if (!test_range_bit(extent_ins, start, end,
EXTENT_WRITEBACK, 0)) {
- mutex_unlock(&info->extent_ins_mutex);
-free_extent:
- ret = __free_extent(trans, extent_root,
- start, end + 1 - start,
- extent_op->orig_parent,
- extent_root->root_key.objectid,
- extent_op->orig_generation,
- extent_op->level, 1, 0);
- kfree(extent_op);
+ list_add_tail(&extent_op->list, &delete_list);
+ nr++;
} else {
kfree(extent_op);
clear_extent_bits(&info->extent_ins, start, end,
EXTENT_WRITEBACK, GFP_NOFS);
- mutex_unlock(&info->extent_ins_mutex);
-
- if (extent_op->type == PENDING_BACKREF_UPDATE)
- goto free_extent;
+ if (extent_op->type == PENDING_BACKREF_UPDATE) {
+ list_add_tail(&extent_op->list, &delete_list);
+ search = end + 1;
+ nr++;
+ continue;
+ }
mutex_lock(&extent_root->fs_info->pinned_mutex);
ret = pin_down_bytes(trans, extent_root, start,
ret = update_block_group(trans, extent_root, start,
end + 1 - start, 0, ret > 0);
+ unlock_extent(extent_ins, start, end, GFP_NOFS);
BUG_ON(ret);
kfree(extent_op);
}
if (ret)
err = ret;
- unlock_extent(extent_ins, start, end, GFP_NOFS);
- if (all)
- search = 0;
- else
- search = end + 1;
- cond_resched();
+ search = end + 1;
+
+ if (need_resched()) {
+ mutex_unlock(&info->extent_ins_mutex);
+ cond_resched();
+ mutex_lock(&info->extent_ins_mutex);
+ }
+ }
+
+ if (nr) {
+ ret = free_extents(trans, extent_root, &delete_list);
+ BUG_ON(ret);
+ }
+
+ if (all && skipped) {
+ INIT_LIST_HEAD(&delete_list);
+ search = 0;
+ nr = 0;
+ goto again;
}
+
return err;
}
WARN_ON(num_bytes < root->sectorsize);
if (root == extent_root) {
- struct pending_extent_op *extent_op;
+ struct pending_extent_op *extent_op = NULL;
+
+ mutex_lock(&root->fs_info->extent_ins_mutex);
+ if (test_range_bit(&root->fs_info->extent_ins, bytenr,
+ bytenr + num_bytes - 1, EXTENT_WRITEBACK, 0)) {
+ u64 priv;
+ ret = get_state_private(&root->fs_info->extent_ins,
+ bytenr, &priv);
+ BUG_ON(ret);
+ extent_op = (struct pending_extent_op *)
+ (unsigned long)priv;
+
+ extent_op->del = 1;
+ if (extent_op->type == PENDING_EXTENT_INSERT) {
+ mutex_unlock(&root->fs_info->extent_ins_mutex);
+ return 0;
+ }
+ }
+
+ if (extent_op) {
+ ref_generation = extent_op->orig_generation;
+ parent = extent_op->orig_parent;
+ }
extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
BUG_ON(!extent_op);
extent_op->generation = ref_generation;
extent_op->orig_generation = ref_generation;
extent_op->level = (int)owner_objectid;
+ INIT_LIST_HEAD(&extent_op->list);
+ extent_op->del = 0;
- mutex_lock(&root->fs_info->extent_ins_mutex);
set_extent_bits(&root->fs_info->pending_del,
bytenr, bytenr + num_bytes - 1,
EXTENT_WRITEBACK, GFP_NOFS);
cache = btrfs_lookup_block_group(root->fs_info, bytenr);
BUG_ON(!cache);
btrfs_add_free_space(cache, bytenr, num_bytes);
+ put_block_group(cache);
update_reserved_extents(root, bytenr, num_bytes, 0);
return 0;
}
* ins->offset == number of blocks
* Any available blocks before search_start are skipped.
*/
-static int noinline find_free_extent(struct btrfs_trans_handle *trans,
+static noinline int find_free_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *orig_root,
u64 num_bytes, u64 empty_size,
u64 search_start, u64 search_end,
int data)
{
int ret = 0;
- struct btrfs_root * root = orig_root->fs_info->extent_root;
+ struct btrfs_root *root = orig_root->fs_info->extent_root;
u64 total_needed = num_bytes;
u64 *last_ptr = NULL;
u64 last_wanted = 0;
int allowed_chunk_alloc = 0;
struct list_head *head = NULL, *cur = NULL;
int loop = 0;
+ int extra_loop = 0;
struct btrfs_space_info *space_info;
WARN_ON(num_bytes < root->sectorsize);
* group thats not of the proper type, while looping this
* should never happen
*/
+ if (empty_size)
+ extra_loop = 1;
+
if (!block_group)
goto new_group_no_lock;
+ if (unlikely(!block_group->cached)) {
+ mutex_lock(&block_group->cache_mutex);
+ ret = cache_block_group(root, block_group);
+ mutex_unlock(&block_group->cache_mutex);
+ if (ret)
+ break;
+ }
+
mutex_lock(&block_group->alloc_mutex);
if (unlikely(!block_group_bits(block_group, data)))
goto new_group;
- ret = cache_block_group(root, block_group);
- if (ret) {
- mutex_unlock(&block_group->alloc_mutex);
- break;
- }
-
- if (block_group->ro)
+ if (unlikely(block_group->ro))
goto new_group;
free_space = btrfs_find_free_space(block_group, search_start,
if (last_wanted && search_start != last_wanted) {
total_needed += empty_cluster;
+ empty_size += empty_cluster;
last_wanted = 0;
/*
* if search_start is still in this block group
}
new_group:
mutex_unlock(&block_group->alloc_mutex);
+ put_block_group(block_group);
+ block_group = NULL;
new_group_no_lock:
+ /* don't try to compare new allocations against the
+ * last allocation any more
+ */
last_wanted = 0;
- if (!allowed_chunk_alloc) {
- total_needed -= empty_size;
- empty_size = 0;
- }
+
/*
* Here's how this works.
* loop == 0: we were searching a block group via a hint
cur = head->next;
loop++;
} else if (loop == 1 && cur == head) {
-
+ int keep_going;
+
+ /* at this point we give up on the empty_size
+ * allocations and just try to allocate the min
+ * space.
+ *
+ * The extra_loop field was set if an empty_size
+ * allocation was attempted above, and if this
+ * is try we need to try the loop again without
+ * the additional empty_size.
+ */
total_needed -= empty_size;
empty_size = 0;
+ keep_going = extra_loop;
+ loop++;
if (allowed_chunk_alloc && !chunk_alloc_done) {
up_read(&space_info->groups_sem);
ret = do_chunk_alloc(trans, root, num_bytes +
2 * 1024 * 1024, data, 1);
- if (ret < 0)
- break;
down_read(&space_info->groups_sem);
- loop++;
+ if (ret < 0)
+ goto loop_check;
head = &space_info->block_groups;
- cur = head->next;
+ /*
+ * we've allocated a new chunk, keep
+ * trying
+ */
+ keep_going = 1;
chunk_alloc_done = 1;
} else if (!allowed_chunk_alloc) {
space_info->force_alloc = 1;
- break;
+ }
+loop_check:
+ if (keep_going) {
+ cur = head->next;
+ extra_loop = 0;
} else {
break;
}
block_group = list_entry(cur, struct btrfs_block_group_cache,
list);
+ atomic_inc(&block_group->count);
+
search_start = block_group->key.objectid;
cur = cur->next;
}
/* we found what we needed */
if (ins->objectid) {
if (!(data & BTRFS_BLOCK_GROUP_DATA))
- trans->block_group = block_group;
+ trans->block_group = block_group->key.objectid;
if (last_ptr)
*last_ptr = ins->objectid + ins->offset;
ret = 0;
} else if (!ret) {
+ printk(KERN_ERR "btrfs searching for %llu bytes, "
+ "num_bytes %llu, loop %d, allowed_alloc %d\n",
+ (unsigned long long)total_needed,
+ (unsigned long long)num_bytes,
+ loop, allowed_chunk_alloc);
ret = -ENOSPC;
}
+ if (block_group)
+ put_block_group(block_group);
up_read(&space_info->groups_sem);
return ret;
struct btrfs_block_group_cache *cache;
struct list_head *l;
- printk(KERN_INFO "space_info has %Lu free, is %sfull\n",
- info->total_bytes - info->bytes_used - info->bytes_pinned -
- info->bytes_reserved, (info->full) ? "" : "not ");
+ printk(KERN_INFO "space_info has %llu free, is %sfull\n",
+ (unsigned long long)(info->total_bytes - info->bytes_used -
+ info->bytes_pinned - info->bytes_reserved),
+ (info->full) ? "" : "not ");
down_read(&info->groups_sem);
list_for_each(l, &info->block_groups) {
cache = list_entry(l, struct btrfs_block_group_cache, list);
spin_lock(&cache->lock);
- printk(KERN_INFO "block group %Lu has %Lu bytes, %Lu used "
- "%Lu pinned %Lu reserved\n",
- cache->key.objectid, cache->key.offset,
- btrfs_block_group_used(&cache->item),
- cache->pinned, cache->reserved);
+ printk(KERN_INFO "block group %llu has %llu bytes, %llu used "
+ "%llu pinned %llu reserved\n",
+ (unsigned long long)cache->key.objectid,
+ (unsigned long long)cache->key.offset,
+ (unsigned long long)btrfs_block_group_used(&cache->item),
+ (unsigned long long)cache->pinned,
+ (unsigned long long)cache->reserved);
btrfs_dump_free_space(cache, bytes);
spin_unlock(&cache->lock);
}
if (data) {
alloc_profile = info->avail_data_alloc_bits &
- info->data_alloc_profile;
+ info->data_alloc_profile;
data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
} else if (root == root->fs_info->chunk_root) {
alloc_profile = info->avail_system_alloc_bits &
- info->system_alloc_profile;
+ info->system_alloc_profile;
data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
} else {
alloc_profile = info->avail_metadata_alloc_bits &
- info->metadata_alloc_profile;
+ info->metadata_alloc_profile;
data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
}
again:
- data = reduce_alloc_profile(root, data);
+ data = btrfs_reduce_alloc_profile(root, data);
/*
* the only place that sets empty_size is btrfs_realloc_node, which
* is not called recursively on allocations
struct btrfs_space_info *sinfo;
sinfo = __find_space_info(root->fs_info, data);
- printk("allocation failed flags %Lu, wanted %Lu\n",
- data, num_bytes);
+ printk(KERN_ERR "btrfs allocation failed flags %llu, "
+ "wanted %llu\n", (unsigned long long)data,
+ (unsigned long long)num_bytes);
dump_space_info(sinfo, num_bytes);
BUG();
}
int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
{
struct btrfs_block_group_cache *cache;
+ int ret = 0;
cache = btrfs_lookup_block_group(root->fs_info, start);
if (!cache) {
- printk(KERN_ERR "Unable to find block group for %Lu\n", start);
+ printk(KERN_ERR "Unable to find block group for %llu\n",
+ (unsigned long long)start);
return -ENOSPC;
}
+
+ ret = btrfs_discard_extent(root, start, len);
+
btrfs_add_free_space(cache, start, len);
+ put_block_group(cache);
update_reserved_extents(root, start, len, 0);
- return 0;
+
+ return ret;
}
int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
parent = ins->objectid;
/* block accounting for super block */
- spin_lock_irq(&info->delalloc_lock);
+ spin_lock(&info->delalloc_lock);
super_used = btrfs_super_bytes_used(&info->super_copy);
btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes);
- spin_unlock_irq(&info->delalloc_lock);
/* block accounting for root item */
root_used = btrfs_root_used(&root->root_item);
btrfs_set_root_used(&root->root_item, root_used + num_bytes);
+ spin_unlock(&info->delalloc_lock);
if (root == extent_root) {
struct pending_extent_op *extent_op;
extent_op->generation = ref_generation;
extent_op->orig_generation = 0;
extent_op->level = (int)owner;
+ INIT_LIST_HEAD(&extent_op->list);
+ extent_op->del = 0;
mutex_lock(&root->fs_info->extent_ins_mutex);
set_extent_bits(&root->fs_info->extent_ins, ins->objectid,
}
update_block:
- ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0);
+ ret = update_block_group(trans, root, ins->objectid,
+ ins->offset, 1, 0);
if (ret) {
- printk("update block group failed for %Lu %Lu\n",
- ins->objectid, ins->offset);
+ printk(KERN_ERR "btrfs update block group failed for %llu "
+ "%llu\n", (unsigned long long)ins->objectid,
+ (unsigned long long)ins->offset);
BUG();
}
out:
struct btrfs_block_group_cache *block_group;
block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
- mutex_lock(&block_group->alloc_mutex);
+ mutex_lock(&block_group->cache_mutex);
cache_block_group(root, block_group);
+ mutex_unlock(&block_group->cache_mutex);
- ret = btrfs_remove_free_space_lock(block_group, ins->objectid,
- ins->offset);
- mutex_unlock(&block_group->alloc_mutex);
+ ret = btrfs_remove_free_space(block_group, ins->objectid,
+ ins->offset);
BUG_ON(ret);
+ put_block_group(block_group);
ret = __btrfs_alloc_reserved_extent(trans, root, parent, root_objectid,
ref_generation, owner, ins);
return ret;
return 0;
}
-static int noinline cache_drop_leaf_ref(struct btrfs_trans_handle *trans,
+static noinline int cache_drop_leaf_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_leaf_ref *ref)
{
return 0;
}
-int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start, u64 len,
- u32 *refs)
+static int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start,
+ u64 len, u32 *refs)
{
int ret;
ret = btrfs_lookup_extent_ref(NULL, root, start, len, refs);
BUG_ON(ret);
-#if 0 // some debugging code in case we see problems here
+#if 0 /* some debugging code in case we see problems here */
/* if the refs count is one, it won't get increased again. But
* if the ref count is > 1, someone may be decreasing it at
* the same time we are.
free_extent_buffer(eb);
}
if (*refs == 1) {
- printk("block %llu went down to one during drop_snap\n",
- (unsigned long long)start);
+ printk(KERN_ERR "btrfs block %llu went down to one "
+ "during drop_snap\n", (unsigned long long)start);
}
}
* helper function for drop_snapshot, this walks down the tree dropping ref
* counts as it goes.
*/
-static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
+static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path, int *level)
{
/*
* walk down to the last node level and free all the leaves
*/
- while(*level >= 0) {
+ while (*level >= 0) {
WARN_ON(*level < 0);
WARN_ON(*level >= BTRFS_MAX_LEVEL);
cur = path->nodes[*level];
*level = 0;
break;
}
- if (printk_ratelimit()) {
- printk("leaf ref miss for bytenr %llu\n",
- (unsigned long long)bytenr);
- }
}
next = btrfs_find_tree_block(root, bytenr, blocksize);
if (!next || !btrfs_buffer_uptodate(next, ptr_gen)) {
* walk_down_tree. The main difference is that it checks reference
* counts while tree blocks are locked.
*/
-static int noinline walk_down_subtree(struct btrfs_trans_handle *trans,
+static noinline int walk_down_subtree(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path, int *level)
{
* to find the first node higher up where we haven't yet gone through
* all the slots
*/
-static int noinline walk_up_tree(struct btrfs_trans_handle *trans,
+static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
int *level, int max_level)
}
}
}
- while(1) {
+ while (1) {
wret = walk_down_tree(trans, root, path, &level);
if (wret > 0)
break;
return min(last, start + nr - 1);
}
-static int noinline relocate_inode_pages(struct inode *inode, u64 start,
+static noinline int relocate_inode_pages(struct inode *inode, u64 start,
u64 len)
{
u64 page_start;
}
set_page_extent_mapped(page);
- btrfs_set_extent_delalloc(inode, page_start, page_end);
if (i == first_index)
set_extent_bits(io_tree, page_start, page_end,
EXTENT_BOUNDARY, GFP_NOFS);
+ btrfs_set_extent_delalloc(inode, page_start, page_end);
set_page_dirty(page);
total_dirty++;
return ret;
}
-static int noinline relocate_data_extent(struct inode *reloc_inode,
+static noinline int relocate_data_extent(struct inode *reloc_inode,
struct btrfs_key *extent_key,
u64 offset)
{
root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
root_objectid == BTRFS_DEV_TREE_OBJECTID ||
- root_objectid == BTRFS_TREE_LOG_OBJECTID)
+ root_objectid == BTRFS_TREE_LOG_OBJECTID ||
+ root_objectid == BTRFS_CSUM_TREE_OBJECTID)
return 1;
return 0;
}
-static int noinline __next_ref_path(struct btrfs_trans_handle *trans,
+static noinline int __next_ref_path(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct btrfs_ref_path *ref_path,
int first_time)
if (level < ref_path->lowest_level)
break;
- if (level >= 0) {
+ if (level >= 0)
bytenr = ref_path->nodes[level];
- } else {
+ else
bytenr = ref_path->extent_start;
- }
BUG_ON(bytenr == 0);
parent = ref_path->nodes[level + 1];
level = ref_path->current_level;
while (level < BTRFS_MAX_LEVEL - 1) {
u64 ref_objectid;
- if (level >= 0) {
+
+ if (level >= 0)
bytenr = ref_path->nodes[level];
- } else {
+ else
bytenr = ref_path->extent_start;
- }
+
BUG_ON(bytenr == 0);
key.objectid = bytenr;
return __next_ref_path(trans, extent_root, ref_path, 0);
}
-static int noinline get_new_locations(struct inode *reloc_inode,
+static noinline int get_new_locations(struct inode *reloc_inode,
struct btrfs_key *extent_key,
u64 offset, int no_fragment,
struct disk_extent **extents,
path->slots[0]++;
}
- WARN_ON(cur_pos + offset > last_byte);
+ BUG_ON(cur_pos + offset > last_byte);
if (cur_pos + offset < last_byte) {
ret = -ENOENT;
goto out;
return ret;
}
-static int noinline replace_one_extent(struct btrfs_trans_handle *trans,
+static noinline int replace_one_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
struct btrfs_key *extent_key,
return 0;
}
-static int noinline invalidate_extent_cache(struct btrfs_root *root,
+static noinline int invalidate_extent_cache(struct btrfs_root *root,
struct extent_buffer *leaf,
struct btrfs_block_group_cache *group,
struct btrfs_root *target_root)
return 0;
}
-static int noinline replace_extents_in_leaf(struct btrfs_trans_handle *trans,
+static noinline int replace_extents_in_leaf(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct extent_buffer *leaf,
struct btrfs_block_group_cache *group,
return 0;
}
-static int noinline init_reloc_tree(struct btrfs_trans_handle *trans,
+static noinline int init_reloc_tree(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
struct btrfs_root *reloc_root;
* tree blocks are shared between reloc trees, so they are also shared
* between subvols.
*/
-static int noinline relocate_one_path(struct btrfs_trans_handle *trans,
+static noinline int relocate_one_path(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
struct btrfs_key *first_key,
else
btrfs_node_key_to_cpu(eb, &keys[level], 0);
}
- if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
+ if (nodes[0] &&
+ ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
eb = path->nodes[0];
ret = replace_extents_in_leaf(trans, reloc_root, eb,
group, reloc_inode);
return 0;
}
-static int noinline relocate_tree_block(struct btrfs_trans_handle *trans,
+static noinline int relocate_tree_block(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
struct btrfs_key *first_key,
return 0;
}
-static int noinline del_extent_zero(struct btrfs_trans_handle *trans,
+static noinline int del_extent_zero(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct btrfs_path *path,
struct btrfs_key *extent_key)
return ret;
}
-static struct btrfs_root noinline *read_ref_root(struct btrfs_fs_info *fs_info,
+static noinline struct btrfs_root *read_ref_root(struct btrfs_fs_info *fs_info,
struct btrfs_ref_path *ref_path)
{
struct btrfs_key root_key;
return btrfs_read_fs_root_no_name(fs_info, &root_key);
}
-static int noinline relocate_one_extent(struct btrfs_root *extent_root,
+static noinline int relocate_one_extent(struct btrfs_root *extent_root,
struct btrfs_path *path,
struct btrfs_key *extent_key,
struct btrfs_block_group_cache *group,
ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS);
if (!ref_path) {
- ret = -ENOMEM;
- goto out;
+ ret = -ENOMEM;
+ goto out;
}
for (loops = 0; ; loops++) {
prev_block = block_start;
}
- if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID &&
- pass >= 2) {
+ btrfs_record_root_in_trans(found_root);
+ if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
+ /*
+ * try to update data extent references while
+ * keeping metadata shared between snapshots.
+ */
+ if (pass == 1) {
+ ret = relocate_one_path(trans, found_root,
+ path, &first_key, ref_path,
+ group, reloc_inode);
+ if (ret < 0)
+ goto out;
+ continue;
+ }
/*
* use fallback method to process the remaining
* references.
if (ret)
goto out;
}
- btrfs_record_root_in_trans(found_root);
ret = replace_one_extent(trans, found_root,
path, extent_key,
&first_key, ref_path,
new_extents, nr_extents);
- if (ret < 0)
- goto out;
- continue;
- }
-
- btrfs_record_root_in_trans(found_root);
- if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
+ } else {
ret = relocate_tree_block(trans, found_root, path,
&first_key, ref_path);
- } else {
- /*
- * try to update data extent references while
- * keeping metadata shared between snapshots.
- */
- ret = relocate_one_path(trans, found_root, path,
- &first_key, ref_path,
- group, reloc_inode);
}
if (ret < 0)
goto out;
u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
- num_devices = root->fs_info->fs_devices->num_devices;
+ num_devices = root->fs_info->fs_devices->rw_devices;
if (num_devices == 1) {
stripped |= BTRFS_BLOCK_GROUP_DUP;
stripped = flags & ~stripped;
return flags;
}
-int __alloc_chunk_for_shrink(struct btrfs_root *root,
+static int __alloc_chunk_for_shrink(struct btrfs_root *root,
struct btrfs_block_group_cache *shrink_block_group,
int force)
{
btrfs_set_inode_generation(leaf, item, 1);
btrfs_set_inode_size(leaf, item, size);
btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
- btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NODATASUM |
- BTRFS_INODE_NOCOMPRESS);
+ btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NOCOMPRESS);
btrfs_mark_buffer_dirty(leaf);
btrfs_release_path(root, path);
out:
return ret;
}
-static struct inode noinline *create_reloc_inode(struct btrfs_fs_info *fs_info,
+static noinline struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
struct btrfs_block_group_cache *group)
{
struct inode *inode = NULL;
} else {
BUG_ON(1);
}
+ BTRFS_I(inode)->index_cnt = group->key.objectid;
err = btrfs_orphan_add(trans, inode);
out:
return inode;
}
+int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len)
+{
+
+ struct btrfs_ordered_sum *sums;
+ struct btrfs_sector_sum *sector_sum;
+ struct btrfs_ordered_extent *ordered;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct list_head list;
+ size_t offset;
+ int ret;
+ u64 disk_bytenr;
+
+ INIT_LIST_HEAD(&list);
+
+ ordered = btrfs_lookup_ordered_extent(inode, file_pos);
+ BUG_ON(ordered->file_offset != file_pos || ordered->len != len);
+
+ disk_bytenr = file_pos + BTRFS_I(inode)->index_cnt;
+ ret = btrfs_lookup_csums_range(root->fs_info->csum_root, disk_bytenr,
+ disk_bytenr + len - 1, &list);
+
+ while (!list_empty(&list)) {
+ sums = list_entry(list.next, struct btrfs_ordered_sum, list);
+ list_del_init(&sums->list);
+
+ sector_sum = sums->sums;
+ sums->bytenr = ordered->start;
+
+ offset = 0;
+ while (offset < sums->len) {
+ sector_sum->bytenr += ordered->start - disk_bytenr;
+ sector_sum++;
+ offset += root->sectorsize;
+ }
+
+ btrfs_add_ordered_sum(inode, ordered, sums);
+ }
+ btrfs_put_ordered_extent(ordered);
+ return 0;
+}
+
int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start)
{
struct btrfs_trans_handle *trans;
block_group = btrfs_lookup_block_group(info, group_start);
BUG_ON(!block_group);
- printk("btrfs relocating block group %llu flags %llu\n",
+ printk(KERN_INFO "btrfs relocating block group %llu flags %llu\n",
(unsigned long long)block_group->key.objectid,
(unsigned long long)block_group->flags);
BUG_ON(IS_ERR(reloc_inode));
__alloc_chunk_for_shrink(root, block_group, 1);
- block_group->ro = 1;
- block_group->space_info->total_bytes -= block_group->key.offset;
+ set_block_group_readonly(block_group);
btrfs_start_delalloc_inodes(info->tree_root);
btrfs_wait_ordered_extents(info->tree_root, 0);
btrfs_remove_leaf_refs(info->tree_root, (u64)-1, 1);
mutex_unlock(&root->fs_info->cleaner_mutex);
- while(1) {
+ while (1) {
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
if (ret < 0)
goto out;
if (pass == 0) {
btrfs_wait_ordered_range(reloc_inode, 0, (u64)-1);
invalidate_mapping_pages(reloc_inode->i_mapping, 0, -1);
- WARN_ON(reloc_inode->i_mapping->nrpages);
}
if (total_found > 0) {
- printk("btrfs found %llu extents in pass %d\n",
+ printk(KERN_INFO "btrfs found %llu extents in pass %d\n",
(unsigned long long)total_found, pass);
pass++;
if (total_found == skipped && pass > 2) {
WARN_ON(block_group->reserved > 0);
WARN_ON(btrfs_block_group_used(&block_group->item) > 0);
spin_unlock(&block_group->lock);
+ put_block_group(block_group);
ret = 0;
out:
btrfs_free_path(path);
return ret;
}
-int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path,
- struct btrfs_key *key)
+static int find_first_block_group(struct btrfs_root *root,
+ struct btrfs_path *path, struct btrfs_key *key)
{
int ret = 0;
struct btrfs_key found_key;
if (ret < 0)
goto out;
- while(1) {
+ while (1) {
slot = path->slots[0];
leaf = path->nodes[0];
if (slot >= btrfs_header_nritems(leaf)) {
down_write(&block_group->space_info->groups_sem);
list_del(&block_group->list);
up_write(&block_group->space_info->groups_sem);
+
+ WARN_ON(atomic_read(&block_group->count) != 1);
kfree(block_group);
spin_lock(&info->block_group_cache_lock);
if (!path)
return -ENOMEM;
- while(1) {
+ while (1) {
ret = find_first_block_group(root, path, &key);
if (ret > 0) {
ret = 0;
break;
}
+ atomic_set(&cache->count, 1);
spin_lock_init(&cache->lock);
mutex_init(&cache->alloc_mutex);
+ mutex_init(&cache->cache_mutex);
INIT_LIST_HEAD(&cache->list);
read_extent_buffer(leaf, &cache->item,
btrfs_item_ptr_offset(leaf, path->slots[0]),
BUG_ON(ret);
set_avail_alloc_bits(root->fs_info, cache->flags);
+ if (btrfs_chunk_readonly(root, cache->key.objectid))
+ set_block_group_readonly(cache);
}
ret = 0;
error:
cache->key.objectid = chunk_offset;
cache->key.offset = size;
+ cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
+ atomic_set(&cache->count, 1);
spin_lock_init(&cache->lock);
mutex_init(&cache->alloc_mutex);
+ mutex_init(&cache->cache_mutex);
INIT_LIST_HEAD(&cache->list);
- btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY);
btrfs_set_block_group_used(&cache->item, bytes_used);
btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
block_group = btrfs_lookup_block_group(root->fs_info, group_start);
BUG_ON(!block_group);
+ BUG_ON(!block_group->ro);
memcpy(&key, &block_group->key, sizeof(key));
list_del(&block_group->list);
up_write(&block_group->space_info->groups_sem);
- /*
- memset(shrink_block_group, 0, sizeof(*shrink_block_group));
- kfree(shrink_block_group);
- */
+ spin_lock(&block_group->space_info->lock);
+ block_group->space_info->total_bytes -= block_group->key.offset;
+ block_group->space_info->bytes_readonly -= block_group->key.offset;
+ spin_unlock(&block_group->space_info->lock);
+ block_group->space_info->full = 0;
+
+ put_block_group(block_group);
+ put_block_group(block_group);
ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
if (ret > 0)
git://ftp.safe.ca / safe / jmp / linux-2.6 / blobdiff