#include "disk-io.h"
#include "extent_io.h"
#include "inode-map.h"
+#include "volumes.h"
#define BITS_PER_BITMAP (PAGE_CACHE_SIZE * 8)
#define MAX_CACHE_BYTES_PER_GIG (32 * 1024)
+struct btrfs_trim_range {
+ u64 start;
+ u64 bytes;
+ struct list_head list;
+};
+
static int link_free_space(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info);
static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
int ret;
struct btrfs_free_cluster *cluster = NULL;
struct rb_node *node = rb_first(&ctl->free_space_offset);
+ struct btrfs_trim_range *trim_entry;
/* Get the cluster for this block_group if it exists */
if (block_group && !list_empty(&block_group->cluster_list)) {
cluster = NULL;
}
}
+
+ /*
+ * Make sure we don't miss any range that was removed from our rbtree
+ * because trimming is running. Otherwise after a umount+mount (or crash
+ * after committing the transaction) we would leak free space and get
+ * an inconsistent free space cache report from fsck.
+ */
+ list_for_each_entry(trim_entry, &ctl->trimming_ranges, list) {
+ ret = io_ctl_add_entry(io_ctl, trim_entry->start,
+ trim_entry->bytes, NULL);
+ if (ret)
+ goto fail;
+ *entries += 1;
+ }
+
return 0;
fail:
return -ENOSPC;
io_ctl_set_generation(&io_ctl, trans->transid);
+ mutex_lock(&ctl->cache_writeout_mutex);
/* Write out the extent entries in the free space cache */
ret = write_cache_extent_entries(&io_ctl, ctl,
block_group, &entries, &bitmaps,
&bitmap_list);
- if (ret)
+ if (ret) {
+ mutex_unlock(&ctl->cache_writeout_mutex);
goto out_nospc;
+ }
/*
* Some spaces that are freed in the current transaction are pinned,
* committed, we shouldn't lose them.
*/
ret = write_pinned_extent_entries(root, block_group, &io_ctl, &entries);
- if (ret)
+ if (ret) {
+ mutex_unlock(&ctl->cache_writeout_mutex);
goto out_nospc;
+ }
- /* At last, we write out all the bitmaps. */
+ /*
+ * At last, we write out all the bitmaps and keep cache_writeout_mutex
+ * locked while doing it because a concurrent trim can be manipulating
+ * or freeing the bitmap.
+ */
ret = write_bitmap_entries(&io_ctl, &bitmap_list);
+ mutex_unlock(&ctl->cache_writeout_mutex);
if (ret)
goto out_nospc;
ctl->start = block_group->key.objectid;
ctl->private = block_group;
ctl->op = &free_space_op;
+ INIT_LIST_HEAD(&ctl->trimming_ranges);
+ mutex_init(&ctl->cache_writeout_mutex);
/*
* we only want to have 32k of ram per block group for keeping
static int do_trimming(struct btrfs_block_group_cache *block_group,
u64 *total_trimmed, u64 start, u64 bytes,
- u64 reserved_start, u64 reserved_bytes)
+ u64 reserved_start, u64 reserved_bytes,
+ struct btrfs_trim_range *trim_entry)
{
struct btrfs_space_info *space_info = block_group->space_info;
struct btrfs_fs_info *fs_info = block_group->fs_info;
+ struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
int ret;
int update = 0;
u64 trimmed = 0;
if (!ret)
*total_trimmed += trimmed;
+ mutex_lock(&ctl->cache_writeout_mutex);
btrfs_add_free_space(block_group, reserved_start, reserved_bytes);
+ list_del(&trim_entry->list);
+ mutex_unlock(&ctl->cache_writeout_mutex);
if (update) {
spin_lock(&space_info->lock);
u64 bytes;
while (start < end) {
+ struct btrfs_trim_range trim_entry;
+
+ mutex_lock(&ctl->cache_writeout_mutex);
spin_lock(&ctl->tree_lock);
if (ctl->free_space < minlen) {
spin_unlock(&ctl->tree_lock);
+ mutex_unlock(&ctl->cache_writeout_mutex);
break;
}
entry = tree_search_offset(ctl, start, 0, 1);
if (!entry) {
spin_unlock(&ctl->tree_lock);
+ mutex_unlock(&ctl->cache_writeout_mutex);
break;
}
node = rb_next(&entry->offset_index);
if (!node) {
spin_unlock(&ctl->tree_lock);
+ mutex_unlock(&ctl->cache_writeout_mutex);
goto out;
}
entry = rb_entry(node, struct btrfs_free_space,
if (entry->offset >= end) {
spin_unlock(&ctl->tree_lock);
+ mutex_unlock(&ctl->cache_writeout_mutex);
break;
}
bytes = min(extent_start + extent_bytes, end) - start;
if (bytes < minlen) {
spin_unlock(&ctl->tree_lock);
+ mutex_unlock(&ctl->cache_writeout_mutex);
goto next;
}
kmem_cache_free(btrfs_free_space_cachep, entry);
spin_unlock(&ctl->tree_lock);
+ trim_entry.start = extent_start;
+ trim_entry.bytes = extent_bytes;
+ list_add_tail(&trim_entry.list, &ctl->trimming_ranges);
+ mutex_unlock(&ctl->cache_writeout_mutex);
ret = do_trimming(block_group, total_trimmed, start, bytes,
- extent_start, extent_bytes);
+ extent_start, extent_bytes, &trim_entry);
if (ret)
break;
next:
while (offset < end) {
bool next_bitmap = false;
+ struct btrfs_trim_range trim_entry;
+ mutex_lock(&ctl->cache_writeout_mutex);
spin_lock(&ctl->tree_lock);
if (ctl->free_space < minlen) {
spin_unlock(&ctl->tree_lock);
+ mutex_unlock(&ctl->cache_writeout_mutex);
break;
}
entry = tree_search_offset(ctl, offset, 1, 0);
if (!entry) {
spin_unlock(&ctl->tree_lock);
+ mutex_unlock(&ctl->cache_writeout_mutex);
next_bitmap = true;
goto next;
}
ret2 = search_bitmap(ctl, entry, &start, &bytes);
if (ret2 || start >= end) {
spin_unlock(&ctl->tree_lock);
+ mutex_unlock(&ctl->cache_writeout_mutex);
next_bitmap = true;
goto next;
}
bytes = min(bytes, end - start);
if (bytes < minlen) {
spin_unlock(&ctl->tree_lock);
+ mutex_unlock(&ctl->cache_writeout_mutex);
goto next;
}
free_bitmap(ctl, entry);
spin_unlock(&ctl->tree_lock);
+ trim_entry.start = start;
+ trim_entry.bytes = bytes;
+ list_add_tail(&trim_entry.list, &ctl->trimming_ranges);
+ mutex_unlock(&ctl->cache_writeout_mutex);
ret = do_trimming(block_group, total_trimmed, start, bytes,
- start, bytes);
+ start, bytes, &trim_entry);
if (ret)
break;
next:
*trimmed = 0;
+ spin_lock(&block_group->lock);
+ if (block_group->removed) {
+ spin_unlock(&block_group->lock);
+ return 0;
+ }
+ atomic_inc(&block_group->trimming);
+ spin_unlock(&block_group->lock);
+
ret = trim_no_bitmap(block_group, trimmed, start, end, minlen);
if (ret)
- return ret;
+ goto out;
ret = trim_bitmaps(block_group, trimmed, start, end, minlen);
+out:
+ spin_lock(&block_group->lock);
+ if (atomic_dec_and_test(&block_group->trimming) &&
+ block_group->removed) {
+ struct extent_map_tree *em_tree;
+ struct extent_map *em;
+
+ spin_unlock(&block_group->lock);
+
+ em_tree = &block_group->fs_info->mapping_tree.map_tree;
+ write_lock(&em_tree->lock);
+ em = lookup_extent_mapping(em_tree, block_group->key.objectid,
+ 1);
+ BUG_ON(!em); /* logic error, can't happen */
+ remove_extent_mapping(em_tree, em);
+ write_unlock(&em_tree->lock);
+
+ lock_chunks(block_group->fs_info->chunk_root);
+ list_del_init(&em->list);
+ unlock_chunks(block_group->fs_info->chunk_root);
+
+ /* once for us and once for the tree */
+ free_extent_map(em);
+ free_extent_map(em);
+
+ /*
+ * We've left one free space entry and other tasks trimming
+ * this block group have left 1 entry each one. Free them.
+ */
+ __btrfs_remove_free_space_cache(block_group->free_space_ctl);
+ } else {
+ spin_unlock(&block_group->lock);
+ }
return ret;
}
projects / cascardo / linux.git / blobdiff