Since we will introduce a new on-disk based dedup method, introduce new
interfaces to resume previous dedup setup.
And since we introduce a new tree for status, also add disable handler
for it.
Signed-off-by: Wang Xiaoguang <wangxg.f...@cn.fujitsu.com>
Signed-off-by: Qu Wenruo <quwen...@cn.fujitsu.com>
---
fs/btrfs/dedup.c | 270 +++++++++++++++++++++++++++++++++++++++++++++++++----
fs/btrfs/dedup.h | 13 +++
fs/btrfs/disk-io.c | 21 ++++-
fs/btrfs/disk-io.h | 1 +
4 files changed, 283 insertions(+), 22 deletions(-)
diff --git a/fs/btrfs/dedup.c b/fs/btrfs/dedup.c
index 9777355..ad7b7e1 100644
--- a/fs/btrfs/dedup.c
+++ b/fs/btrfs/dedup.c
@@ -21,6 +21,8 @@
#include "transaction.h"
#include "delayed-ref.h"
#include "qgroup.h"
+#include "disk-io.h"
+#include "locking.h"
struct inmem_hash {
struct rb_node hash_node;
@@ -41,10 +43,103 @@ static inline struct inmem_hash *inmem_alloc_hash(u16 type)
GFP_NOFS);
}
+static int init_dedup_info(struct btrfs_dedup_info **ret_info, u16 type,
+ u16 backend, u64 blocksize, u64 limit)
+{
+ struct btrfs_dedup_info *dedup_info;
+
+ dedup_info = kzalloc(sizeof(*dedup_info), GFP_NOFS);
+ if (!dedup_info)
+ return -ENOMEM;
+
+ dedup_info->hash_type = type;
+ dedup_info->backend = backend;
+ dedup_info->blocksize = blocksize;
+ dedup_info->limit_nr = limit;
+
+ /* only support SHA256 yet */
+ dedup_info->dedup_driver = crypto_alloc_shash("sha256", 0, 0);
+ if (IS_ERR(dedup_info->dedup_driver)) {
+ int ret;
+
+ ret = PTR_ERR(dedup_info->dedup_driver);
+ kfree(dedup_info);
+ return ret;
+ }
+
+ dedup_info->hash_root = RB_ROOT;
+ dedup_info->bytenr_root = RB_ROOT;
+ dedup_info->current_nr = 0;
+ INIT_LIST_HEAD(&dedup_info->lru_list);
+ mutex_init(&dedup_info->lock);
+
+ *ret_info = dedup_info;
+ return 0;
+}
+
+static int init_dedup_tree(struct btrfs_fs_info *fs_info,
+ struct btrfs_dedup_info *dedup_info)
+{
+ struct btrfs_root *dedup_root;
+ struct btrfs_key key;
+ struct btrfs_path *path;
+ struct btrfs_dedup_status_item *status;
+ struct btrfs_trans_handle *trans;
+ int ret;
+
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
+
+ trans = btrfs_start_transaction(fs_info->tree_root, 2);
+ if (IS_ERR(trans)) {
+ ret = PTR_ERR(trans);
+ goto out;
+ }
+ dedup_root = btrfs_create_tree(trans, fs_info,
+ BTRFS_DEDUP_TREE_OBJECTID);
+ if (IS_ERR(dedup_root)) {
+ ret = PTR_ERR(dedup_root);
+ btrfs_abort_transaction(trans, fs_info->tree_root, ret);
+ goto out;
+ }
+ dedup_info->dedup_root = dedup_root;
+
+ key.objectid = 0;
+ key.type = BTRFS_DEDUP_STATUS_ITEM_KEY;
+ key.offset = 0;
+
+ ret = btrfs_insert_empty_item(trans, dedup_root, path, &key,
+ sizeof(*status));
+ if (ret < 0) {
+ btrfs_abort_transaction(trans, fs_info->tree_root, ret);
+ goto out;
+ }
+
+ status = btrfs_item_ptr(path->nodes[0], path->slots[0],
+ struct btrfs_dedup_status_item);
+ btrfs_set_dedup_status_blocksize(path->nodes[0], status,
+ dedup_info->blocksize);
+ btrfs_set_dedup_status_limit(path->nodes[0], status,
+ dedup_info->limit_nr);
+ btrfs_set_dedup_status_hash_type(path->nodes[0], status,
+ dedup_info->hash_type);
+ btrfs_set_dedup_status_backend(path->nodes[0], status,
+ dedup_info->backend);
+ btrfs_mark_buffer_dirty(path->nodes[0]);
+out:
+ btrfs_free_path(path);
+ if (ret == 0)
+ btrfs_commit_transaction(trans, fs_info->tree_root);
+ return ret;
+}
+
int btrfs_dedup_enable(struct btrfs_fs_info *fs_info, u16 type, u16 backend,
u64 blocksize, u64 limit_nr)
{
struct btrfs_dedup_info *dedup_info;
+ int create_tree;
+ u64 compat_ro_flag = btrfs_super_compat_ro_flags(fs_info->super_copy);
u64 limit = limit_nr;
int ret = 0;
@@ -63,6 +158,14 @@ int btrfs_dedup_enable(struct btrfs_fs_info *fs_info, u16
type, u16 backend,
limit = 4096; /* default value */
if (backend == BTRFS_DEDUP_BACKEND_ONDISK && limit_nr != 0)
limit = 0;
+ /* Ondisk backend needs DEDUP RO compat feature */
+ if (!(compat_ro_flag & BTRFS_FEATURE_COMPAT_RO_DEDUP) &&
+ backend == BTRFS_DEDUP_BACKEND_ONDISK)
+ return -EOPNOTSUPP;
+
+ /* Meaningless and unable to enable dedup for RO fs */
+ if (fs_info->sb->s_flags & MS_RDONLY)
+ return -EROFS;
dedup_info = fs_info->dedup_info;
if (dedup_info) {
@@ -80,30 +183,71 @@ int btrfs_dedup_enable(struct btrfs_fs_info *fs_info, u16
type, u16 backend,
mutex_unlock(&dedup_info->lock);
return 0;
}
-
+ dedup_info = NULL;
enable:
- dedup_info = kzalloc(sizeof(*dedup_info), GFP_NOFS);
- if (dedup_info)
+ create_tree = compat_ro_flag & BTRFS_FEATURE_COMPAT_RO_DEDUP;
+
+ ret = init_dedup_info(&dedup_info, type, backend, blocksize, limit);
+ if (ret < 0)
+ return ret;
+ if (create_tree) {
+ ret = init_dedup_tree(fs_info, dedup_info);
+ if (ret < 0)
+ goto out;
+ }
+
+ fs_info->dedup_info = dedup_info;
+ /* We must ensure dedup_enabled is modified after dedup_info */
+ smp_wmb();
+ fs_info->dedup_enabled = 1;
+out:
+ if (ret < 0) {
+ crypto_free_shash(dedup_info->dedup_driver);
+ kfree(dedup_info);
+ }
+ return ret;
+}
+
+int btrfs_dedup_resume(struct btrfs_fs_info *fs_info,
+ struct btrfs_root *dedup_root)
+{
+ struct btrfs_dedup_info *dedup_info;
+ struct btrfs_dedup_status_item *status;
+ struct btrfs_key key;
+ struct btrfs_path *path;
+ u64 blocksize;
+ u64 limit;
+ u16 type;
+ u16 backend;
+ int ret = 0;
+
+ path = btrfs_alloc_path();
+ if (!path)
return -ENOMEM;
- dedup_info->hash_type = type;
- dedup_info->backend = backend;
- dedup_info->blocksize = blocksize;
- dedup_info->limit_nr = limit;
+ key.objectid = 0;
+ key.type = BTRFS_DEDUP_STATUS_ITEM_KEY;
+ key.offset = 0;
- /* Only support SHA256 yet */
- dedup_info->dedup_driver = crypto_alloc_shash("sha256", 0, 0);
- if (IS_ERR(dedup_info->dedup_driver)) {
- btrfs_err(fs_info, "failed to init sha256 driver");
- ret = PTR_ERR(dedup_info->dedup_driver);
+ ret = btrfs_search_slot(NULL, dedup_root, &key, path, 0, 0);
+ if (ret > 0) {
+ ret = -ENOENT;
+ goto out;
+ } else if (ret < 0) {
goto out;
}
- dedup_info->hash_root = RB_ROOT;
- dedup_info->bytenr_root = RB_ROOT;
- dedup_info->current_nr = 0;
- INIT_LIST_HEAD(&dedup_info->lru_list);
- mutex_init(&dedup_info->lock);
+ status = btrfs_item_ptr(path->nodes[0], path->slots[0],
+ struct btrfs_dedup_status_item);
+ blocksize = btrfs_dedup_status_blocksize(path->nodes[0], status);
+ limit = btrfs_dedup_status_limit(path->nodes[0], status);
+ type = btrfs_dedup_status_hash_type(path->nodes[0], status);
+ backend = btrfs_dedup_status_backend(path->nodes[0], status);
+
+ ret = init_dedup_info(&dedup_info, type, backend, blocksize, limit);
+ if (ret < 0)
+ goto out;
+ dedup_info->dedup_root = dedup_root;
fs_info->dedup_info = dedup_info;
/* We must ensure dedup_enabled is modified after dedup_info */
@@ -111,11 +255,36 @@ enable:
fs_info->dedup_enabled = 1;
out:
- if (ret < 0)
- kfree(dedup_info);
+ btrfs_free_path(path);
return ret;
}
+static void inmem_destroy(struct btrfs_dedup_info *dedup_info);
+int btrfs_dedup_cleanup(struct btrfs_fs_info *fs_info)
+{
+ struct btrfs_dedup_info *dedup_info;
+
+ fs_info->dedup_enabled = 0;
+
+ smp_wmb();
+ /* same as disable */
+ dedup_info = fs_info->dedup_info;
+ fs_info->dedup_info = NULL;
+
+ if (!dedup_info)
+ return 0;
+
+ if (dedup_info->backend == BTRFS_DEDUP_BACKEND_INMEMORY)
+ inmem_destroy(dedup_info);
+ if (dedup_info->dedup_root) {
+ free_root_extent_buffers(dedup_info->dedup_root);
+ kfree(dedup_info->dedup_root);
+ }
+ crypto_free_shash(dedup_info->dedup_driver);
+ kfree(dedup_info);
+ return 0;
+}
+
static int inmem_insert_hash(struct rb_root *root,
struct inmem_hash *hash, int hash_len)
{
@@ -325,6 +494,65 @@ static void inmem_destroy(struct btrfs_dedup_info
*dedup_info)
mutex_unlock(&dedup_info->lock);
}
+static int remove_dedup_tree(struct btrfs_root *dedup_root)
+{
+ struct btrfs_trans_handle *trans;
+ struct btrfs_fs_info *fs_info = dedup_root->fs_info;
+ struct btrfs_path *path;
+ struct btrfs_key key;
+ struct extent_buffer *node;
+ int ret;
+ int nr;
+
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
+ trans = btrfs_start_transaction(fs_info->tree_root, 2);
+ if (IS_ERR(trans)) {
+ ret = PTR_ERR(trans);
+ goto out;
+ }
+
+ path->leave_spinning = 1;
+ key.objectid = 0;
+ key.offset = 0;
+ key.type = 0;
+
+ while (1) {
+ ret = btrfs_search_slot(trans, dedup_root, &key, path, -1, 1);
+ if (ret < 0)
+ goto out;
+ node = path->nodes[0];
+ nr = btrfs_header_nritems(node);
+ if (nr == 0) {
+ btrfs_release_path(path);
+ break;
+ }
+ path->slots[0] = 0;
+ ret = btrfs_del_items(trans, dedup_root, path, 0, nr);
+ if (ret)
+ goto out;
+ btrfs_release_path(path);
+ }
+
+ ret = btrfs_del_root(trans, fs_info->tree_root, &dedup_root->root_key);
+ if (ret)
+ goto out;
+
+ list_del(&dedup_root->dirty_list);
+ btrfs_tree_lock(dedup_root->node);
+ clean_tree_block(trans, fs_info, dedup_root->node);
+ btrfs_tree_unlock(dedup_root->node);
+ btrfs_free_tree_block(trans, dedup_root, dedup_root->node, 0, 1);
+ free_extent_buffer(dedup_root->node);
+ free_extent_buffer(dedup_root->commit_root);
+ kfree(dedup_root);
+ ret = btrfs_commit_transaction(trans, fs_info->tree_root);
+out:
+ btrfs_free_path(path);
+ return ret;
+}
+
int btrfs_dedup_disable(struct btrfs_fs_info *fs_info)
{
struct btrfs_dedup_info *dedup_info;
@@ -358,10 +586,12 @@ int btrfs_dedup_disable(struct btrfs_fs_info *fs_info)
/* now we are OK to clean up everything */
if (dedup_info->backend == BTRFS_DEDUP_BACKEND_INMEMORY)
inmem_destroy(dedup_info);
+ if (dedup_info->dedup_root)
+ ret = remove_dedup_tree(dedup_info->dedup_root);
crypto_free_shash(dedup_info->dedup_driver);
kfree(dedup_info);
- return 0;
+ return ret;
}
/*
diff --git a/fs/btrfs/dedup.h b/fs/btrfs/dedup.h
index 4f681bb..31c6b17 100644
--- a/fs/btrfs/dedup.h
+++ b/fs/btrfs/dedup.h
@@ -109,6 +109,19 @@ int btrfs_dedup_enable(struct btrfs_fs_info *fs_info, u16
type, u16 backend,
int btrfs_dedup_disable(struct btrfs_fs_info *fs_info);
/*
+ * Restore previous dedup setup from disk
+ * Called at mount time
+ */
+int btrfs_dedup_resume(struct btrfs_fs_info *fs_info,
+ struct btrfs_root *dedup_root);
+
+/*
+ * Cleanup current btrfs_dedup_info
+ * Called in umount time
+ */
+int btrfs_dedup_cleanup(struct btrfs_fs_info *fs_info);
+
+/*
* Calculate hash for dedup.
* Caller must ensure [start, start + dedup_bs) has valid data.
*/
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index baefe33..23cb76a 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -50,6 +50,7 @@
#include "raid56.h"
#include "sysfs.h"
#include "qgroup.h"
+#include "dedup.h"
#ifdef CONFIG_X86
#include <asm/cpufeature.h>
@@ -2156,7 +2157,7 @@ static void btrfs_stop_all_workers(struct btrfs_fs_info
*fs_info)
btrfs_destroy_workqueue(fs_info->extent_workers);
}
-static void free_root_extent_buffers(struct btrfs_root *root)
+void free_root_extent_buffers(struct btrfs_root *root)
{
if (root) {
free_extent_buffer(root->node);
@@ -2488,7 +2489,21 @@ static int btrfs_read_roots(struct btrfs_fs_info
*fs_info,
fs_info->free_space_root = root;
}
- return 0;
+ location.objectid = BTRFS_DEDUP_TREE_OBJECTID;
+ root = btrfs_read_tree_root(tree_root, &location);
+ if (IS_ERR(root)) {
+ ret = PTR_ERR(root);
+ if (ret != -ENOENT)
+ return ret;
+ return 0;
+ }
+ set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
+ ret = btrfs_dedup_resume(fs_info, root);
+ if (ret < 0) {
+ free_root_extent_buffers(root);
+ kfree(root);
+ }
+ return ret;
}
int open_ctree(struct super_block *sb,
@@ -3883,6 +3898,8 @@ void close_ctree(struct btrfs_root *root)
btrfs_free_qgroup_config(fs_info);
+ btrfs_dedup_cleanup(fs_info);
+
if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
btrfs_info(fs_info, "at unmount delalloc count %lld",
percpu_counter_sum(&fs_info->delalloc_bytes));
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 8e79d00..42c4ff2 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -70,6 +70,7 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_root
*tree_root,
int btrfs_init_fs_root(struct btrfs_root *root);
int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
struct btrfs_root *root);
+void free_root_extent_buffers(struct btrfs_root *root);
void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info);
struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
--
2.7.1
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
