From 9bf7a4890518186238d2579be16ecc5190a707c0 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Fri, 1 Mar 2013 13:35:47 -0500 Subject: Btrfs: use set_nlink if our i_nlink is 0 We need to inc the nlink of deleted entries when running replay so we can do the unlink on the fs_root and get everything cleaned up and then have the orphan cleanup do the right thing. The problem is inc_nlink complains about this, even thought it still does the right thing. So use set_nlink() if our i_nlink is 0 to keep users from seeing the warnings during log replay. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/tree-log.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index c7ef569..451fad9 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -1382,7 +1382,10 @@ static noinline int link_to_fixup_dir(struct btrfs_trans_handle *trans, btrfs_release_path(path); if (ret == 0) { - btrfs_inc_nlink(inode); + if (!inode->i_nlink) + set_nlink(inode, 1); + else + btrfs_inc_nlink(inode); ret = btrfs_update_inode(trans, root, inode); } else if (ret == -EEXIST) { ret = 0; -- cgit v1.1 From aec8030a8745221c8658f2033b22c98528897b13 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Mon, 4 Mar 2013 09:44:29 +0000 Subject: Btrfs: fix wrong handle at error path of create_snapshot() when the commit fails There are several bugs at error path of create_snapshot() when the transaction commitment failed. - access the freed transaction handler. At the end of the transaction commitment, the transaction handler was freed, so we should not access it after the transaction commitment. - we were not aware of the error which happened during the snapshot creation if we submitted a async transaction commitment. - pending snapshot access vs pending snapshot free. when something wrong happened after we submitted a async transaction commitment, the transaction committer would cleanup the pending snapshots and free them. But the snapshot creators were not aware of it, they would access the freed pending snapshots. This patch fixes the above problems by: - remove the dangerous code that accessed the freed handler - assign ->error if the error happens during the snapshot creation - the transaction committer doesn't free the pending snapshots, just assigns the error number and evicts them before we unblock the transaction. Reported-by: Dan Carpenter Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/disk-io.c | 16 ++++++-------- fs/btrfs/ioctl.c | 6 +----- fs/btrfs/transaction.c | 58 ++++++++++++++++++++++++++++---------------------- 3 files changed, 41 insertions(+), 39 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 02369a3..7d84651 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -62,7 +62,7 @@ static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, static void btrfs_destroy_ordered_extents(struct btrfs_root *root); static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, struct btrfs_root *root); -static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t); +static void btrfs_evict_pending_snapshots(struct btrfs_transaction *t); static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); static int btrfs_destroy_marked_extents(struct btrfs_root *root, struct extent_io_tree *dirty_pages, @@ -3687,7 +3687,7 @@ int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, return ret; } -static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t) +static void btrfs_evict_pending_snapshots(struct btrfs_transaction *t) { struct btrfs_pending_snapshot *snapshot; struct list_head splice; @@ -3700,10 +3700,8 @@ static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t) snapshot = list_entry(splice.next, struct btrfs_pending_snapshot, list); - + snapshot->error = -ECANCELED; list_del_init(&snapshot->list); - - kfree(snapshot); } } @@ -3840,6 +3838,8 @@ void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, cur_trans->blocked = 1; wake_up(&root->fs_info->transaction_blocked_wait); + btrfs_evict_pending_snapshots(cur_trans); + cur_trans->blocked = 0; wake_up(&root->fs_info->transaction_wait); @@ -3849,8 +3849,6 @@ void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, btrfs_destroy_delayed_inodes(root); btrfs_assert_delayed_root_empty(root); - btrfs_destroy_pending_snapshots(cur_trans); - btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, EXTENT_DIRTY); btrfs_destroy_pinned_extent(root, @@ -3894,6 +3892,8 @@ int btrfs_cleanup_transaction(struct btrfs_root *root) if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) wake_up(&root->fs_info->transaction_blocked_wait); + btrfs_evict_pending_snapshots(t); + t->blocked = 0; smp_mb(); if (waitqueue_active(&root->fs_info->transaction_wait)) @@ -3907,8 +3907,6 @@ int btrfs_cleanup_transaction(struct btrfs_root *root) btrfs_destroy_delayed_inodes(root); btrfs_assert_delayed_root_empty(root); - btrfs_destroy_pending_snapshots(t); - btrfs_destroy_delalloc_inodes(root); spin_lock(&root->fs_info->trans_lock); diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index b908960..94c0e42 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -596,12 +596,8 @@ static int create_snapshot(struct btrfs_root *root, struct inode *dir, ret = btrfs_commit_transaction(trans, root->fs_info->extent_root); } - if (ret) { - /* cleanup_transaction has freed this for us */ - if (trans->aborted) - pending_snapshot = NULL; + if (ret) goto fail; - } ret = pending_snapshot->error; if (ret) diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index f11c2e0..d8fce6f 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1053,7 +1053,12 @@ int btrfs_defrag_root(struct btrfs_root *root) /* * new snapshots need to be created at a very specific time in the - * transaction commit. This does the actual creation + * transaction commit. This does the actual creation. + * + * Note: + * If the error which may affect the commitment of the current transaction + * happens, we should return the error number. If the error which just affect + * the creation of the pending snapshots, just return 0. */ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info, @@ -1072,7 +1077,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, struct extent_buffer *tmp; struct extent_buffer *old; struct timespec cur_time = CURRENT_TIME; - int ret; + int ret = 0; u64 to_reserve = 0; u64 index = 0; u64 objectid; @@ -1081,40 +1086,36 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); if (!path) { - ret = pending->error = -ENOMEM; - return ret; + pending->error = -ENOMEM; + return 0; } new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS); if (!new_root_item) { - ret = pending->error = -ENOMEM; + pending->error = -ENOMEM; goto root_item_alloc_fail; } - ret = btrfs_find_free_objectid(tree_root, &objectid); - if (ret) { - pending->error = ret; + pending->error = btrfs_find_free_objectid(tree_root, &objectid); + if (pending->error) goto no_free_objectid; - } btrfs_reloc_pre_snapshot(trans, pending, &to_reserve); if (to_reserve > 0) { - ret = btrfs_block_rsv_add(root, &pending->block_rsv, - to_reserve, - BTRFS_RESERVE_NO_FLUSH); - if (ret) { - pending->error = ret; + pending->error = btrfs_block_rsv_add(root, + &pending->block_rsv, + to_reserve, + BTRFS_RESERVE_NO_FLUSH); + if (pending->error) goto no_free_objectid; - } } - ret = btrfs_qgroup_inherit(trans, fs_info, root->root_key.objectid, - objectid, pending->inherit); - if (ret) { - pending->error = ret; + pending->error = btrfs_qgroup_inherit(trans, fs_info, + root->root_key.objectid, + objectid, pending->inherit); + if (pending->error) goto no_free_objectid; - } key.objectid = objectid; key.offset = (u64)-1; @@ -1142,7 +1143,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, dentry->d_name.len, 0); if (dir_item != NULL && !IS_ERR(dir_item)) { pending->error = -EEXIST; - goto fail; + goto dir_item_existed; } else if (IS_ERR(dir_item)) { ret = PTR_ERR(dir_item); btrfs_abort_transaction(trans, root, ret); @@ -1273,6 +1274,8 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, if (ret) btrfs_abort_transaction(trans, root, ret); fail: + pending->error = ret; +dir_item_existed: trans->block_rsv = rsv; trans->bytes_reserved = 0; no_free_objectid: @@ -1288,12 +1291,17 @@ root_item_alloc_fail: static noinline int create_pending_snapshots(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info) { - struct btrfs_pending_snapshot *pending; + struct btrfs_pending_snapshot *pending, *next; struct list_head *head = &trans->transaction->pending_snapshots; + int ret = 0; - list_for_each_entry(pending, head, list) - create_pending_snapshot(trans, fs_info, pending); - return 0; + list_for_each_entry_safe(pending, next, head, list) { + list_del(&pending->list); + ret = create_pending_snapshot(trans, fs_info, pending); + if (ret) + break; + } + return ret; } static void update_super_roots(struct btrfs_root *root) -- cgit v1.1 From 00d71c9c17b1fd28fa54f323a29a0e23c6d3de40 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Mon, 4 Mar 2013 09:45:06 +0000 Subject: Btrfs: fix unclosed transaction handler when the async transaction commitment fails If the async transaction commitment failed, we need close the current transaction handler, or the current transaction will be blocked to commit because of this orphan handler. We fix the problem by doing sync transaction commitment, that is to invoke btrfs_commit_transaction(). Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/ioctl.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'fs') diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 94c0e42..3fdfabc 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -527,6 +527,8 @@ fail: if (async_transid) { *async_transid = trans->transid; err = btrfs_commit_transaction_async(trans, root, 1); + if (err) + err = btrfs_commit_transaction(trans, root); } else { err = btrfs_commit_transaction(trans, root); } @@ -592,6 +594,8 @@ static int create_snapshot(struct btrfs_root *root, struct inode *dir, *async_transid = trans->transid; ret = btrfs_commit_transaction_async(trans, root->fs_info->extent_root, 1); + if (ret) + ret = btrfs_commit_transaction(trans, root); } else { ret = btrfs_commit_transaction(trans, root->fs_info->extent_root); -- cgit v1.1 From 8f71f3e0e4fb5a2445fb93d3057a33aefc4aa30d Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Mon, 4 Mar 2013 16:25:36 +0000 Subject: Btrfs: check for NULL pointer in updating reloc roots Add a check for NULL pointer to avoid invalid reference. Signed-off-by: Liu Bo Signed-off-by: Josef Bacik --- fs/btrfs/relocation.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs') diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index ba5a321..16e0c6f 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -1269,6 +1269,8 @@ static int __update_reloc_root(struct btrfs_root *root, int del) } spin_unlock(&rc->reloc_root_tree.lock); + if (!node) + return 0; BUG_ON((struct btrfs_root *)node->data != root); if (!del) { -- cgit v1.1 From aca1bba6f9bc550a33312b28e98b3de5ddb3bb15 Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Mon, 4 Mar 2013 16:25:37 +0000 Subject: Btrfs: build up error handling for merge_reloc_roots We first use btrfs_std_error hook to replace with BUG_ON, and we also need to cleanup what is left, including reloc roots rbtree and reloc roots list. Here we use a helper function to cleanup both rbtree and list, and since this function can also be used in the balance recover path, we also make the change as well to keep code simple. Signed-off-by: Liu Bo Signed-off-by: Josef Bacik --- fs/btrfs/relocation.c | 47 +++++++++++++++++++++++++++++++++++------------ 1 file changed, 35 insertions(+), 12 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index 16e0c6f..0f001c1 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -2240,13 +2240,28 @@ again: } static noinline_for_stack +void free_reloc_roots(struct list_head *list) +{ + struct btrfs_root *reloc_root; + + while (!list_empty(list)) { + reloc_root = list_entry(list->next, struct btrfs_root, + root_list); + __update_reloc_root(reloc_root, 1); + free_extent_buffer(reloc_root->node); + free_extent_buffer(reloc_root->commit_root); + kfree(reloc_root); + } +} + +static noinline_for_stack int merge_reloc_roots(struct reloc_control *rc) { struct btrfs_root *root; struct btrfs_root *reloc_root; LIST_HEAD(reloc_roots); int found = 0; - int ret; + int ret = 0; again: root = rc->extent_root; @@ -2272,20 +2287,33 @@ again: BUG_ON(root->reloc_root != reloc_root); ret = merge_reloc_root(rc, root); - BUG_ON(ret); + if (ret) + goto out; } else { list_del_init(&reloc_root->root_list); } ret = btrfs_drop_snapshot(reloc_root, rc->block_rsv, 0, 1); - BUG_ON(ret < 0); + if (ret < 0) { + if (list_empty(&reloc_root->root_list)) + list_add_tail(&reloc_root->root_list, + &reloc_roots); + goto out; + } } if (found) { found = 0; goto again; } +out: + if (ret) { + btrfs_std_error(root->fs_info, ret); + if (!list_empty(&reloc_roots)) + free_reloc_roots(&reloc_roots); + } + BUG_ON(!RB_EMPTY_ROOT(&rc->reloc_root_tree.rb_root)); - return 0; + return ret; } static void free_block_list(struct rb_root *blocks) @@ -4266,14 +4294,9 @@ int btrfs_recover_relocation(struct btrfs_root *root) out_free: kfree(rc); out: - while (!list_empty(&reloc_roots)) { - reloc_root = list_entry(reloc_roots.next, - struct btrfs_root, root_list); - list_del(&reloc_root->root_list); - free_extent_buffer(reloc_root->node); - free_extent_buffer(reloc_root->commit_root); - kfree(reloc_root); - } + if (!list_empty(&reloc_roots)) + free_reloc_roots(&reloc_roots); + btrfs_free_path(path); if (err == 0) { -- cgit v1.1 From e1a1267054ea9dcafd01636e52d441f809efad5e Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Mon, 4 Mar 2013 16:25:38 +0000 Subject: Btrfs: free all recorded tree blocks on error We've missed the 'free blocks' part on ENOMEM error. Signed-off-by: Liu Bo Signed-off-by: Josef Bacik --- fs/btrfs/relocation.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index 0f001c1..7d1654c 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -2848,8 +2848,10 @@ int relocate_tree_blocks(struct btrfs_trans_handle *trans, int err = 0; path = btrfs_alloc_path(); - if (!path) - return -ENOMEM; + if (!path) { + err = -ENOMEM; + goto out_path; + } rb_node = rb_first(blocks); while (rb_node) { @@ -2888,10 +2890,11 @@ int relocate_tree_blocks(struct btrfs_trans_handle *trans, rb_node = rb_next(rb_node); } out: - free_block_list(blocks); err = finish_pending_nodes(trans, rc, path, err); btrfs_free_path(path); +out_path: + free_block_list(blocks); return err; } -- cgit v1.1 From 288189471d7e12fb22c37870e151833f438deea8 Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Mon, 4 Mar 2013 16:25:39 +0000 Subject: Btrfs: do not BUG_ON in prepare_to_reloc We can bail out from here gracefully instead of a cold BUG_ON. Signed-off-by: Liu Bo Signed-off-by: Josef Bacik --- fs/btrfs/relocation.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index 7d1654c..9d13786 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -3731,7 +3731,15 @@ int prepare_to_relocate(struct reloc_control *rc) set_reloc_control(rc); trans = btrfs_join_transaction(rc->extent_root); - BUG_ON(IS_ERR(trans)); + if (IS_ERR(trans)) { + unset_reloc_control(rc); + /* + * extent tree is not a ref_cow tree and has no reloc_root to + * cleanup. And callers are responsible to free the above + * block rsv. + */ + return PTR_ERR(trans); + } btrfs_commit_transaction(trans, rc->extent_root); return 0; } -- cgit v1.1 From 0f788c58194e4ccc5b3ab23f872c5e18542335e4 Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Mon, 4 Mar 2013 16:25:40 +0000 Subject: Btrfs: do not BUG_ON on aborted situation Btrfs balance can easily hit BUG_ON in these places, but we want to it bail out gracefully after we force the whole filesystem to readonly. So we use btrfs_std_error hook in place of BUG_ON. Signed-off-by: Liu Bo Signed-off-by: Josef Bacik --- fs/btrfs/relocation.c | 6 +++++- fs/btrfs/volumes.c | 9 +++++++-- 2 files changed, 12 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index 9d13786..3ebe8797 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -3771,7 +3771,11 @@ static noinline_for_stack int relocate_block_group(struct reloc_control *rc) while (1) { progress++; trans = btrfs_start_transaction(rc->extent_root, 0); - BUG_ON(IS_ERR(trans)); + if (IS_ERR(trans)) { + err = PTR_ERR(trans); + trans = NULL; + break; + } restart: if (update_backref_cache(trans, &rc->backref_cache)) { btrfs_end_transaction(trans, rc->extent_root); diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 35bb2d4..9ff454d 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -2379,7 +2379,11 @@ static int btrfs_relocate_chunk(struct btrfs_root *root, return ret; trans = btrfs_start_transaction(root, 0); - BUG_ON(IS_ERR(trans)); + if (IS_ERR(trans)) { + ret = PTR_ERR(trans); + btrfs_std_error(root->fs_info, ret); + return ret; + } lock_chunks(root); @@ -3050,7 +3054,8 @@ static void __cancel_balance(struct btrfs_fs_info *fs_info) unset_balance_control(fs_info); ret = del_balance_item(fs_info->tree_root); - BUG_ON(ret); + if (ret) + btrfs_std_error(fs_info, ret); atomic_set(&fs_info->mutually_exclusive_operation_running, 0); } -- cgit v1.1 From 66b6135b7cf741f6f44ba938b27583ea3b83bd12 Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Mon, 4 Mar 2013 16:25:41 +0000 Subject: Btrfs: avoid deadlock on transaction waiting list Only let one trans handle to wait for other handles, otherwise we will get ABBA issues. Signed-off-by: Liu Bo Signed-off-by: Josef Bacik --- fs/btrfs/transaction.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'fs') diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index d8fce6f..fedede1 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1457,6 +1457,13 @@ static void cleanup_transaction(struct btrfs_trans_handle *trans, btrfs_abort_transaction(trans, root, err); spin_lock(&root->fs_info->trans_lock); + + if (list_empty(&cur_trans->list)) { + spin_unlock(&root->fs_info->trans_lock); + btrfs_end_transaction(trans, root); + return; + } + list_del_init(&cur_trans->list); if (cur_trans == root->fs_info->running_transaction) { root->fs_info->trans_no_join = 1; -- cgit v1.1 From 9b53157aac7366cea413ee29b629f83225829e87 Mon Sep 17 00:00:00 2001 From: Stefan Behrens Date: Mon, 4 Mar 2013 17:28:38 +0000 Subject: Btrfs: allow running defrag in parallel to administrative tasks Commit 5ac00add added a testnset mutex and code that disallows running administrative tasks in parallel. It is prevented that the device add/delete/balance/replace/resize operations are started in parallel. By mistake, the defragmentation operation was included in the check for mutually exclusiveness as well. This is fixed with this commit. Signed-off-by: Stefan Behrens Signed-off-by: Josef Bacik --- fs/btrfs/ioctl.c | 8 -------- 1 file changed, 8 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 3fdfabc..898c572 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -2245,13 +2245,6 @@ static int btrfs_ioctl_defrag(struct file *file, void __user *argp) if (ret) return ret; - if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, - 1)) { - pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n"); - mnt_drop_write_file(file); - return -EINVAL; - } - if (btrfs_root_readonly(root)) { ret = -EROFS; goto out; @@ -2306,7 +2299,6 @@ static int btrfs_ioctl_defrag(struct file *file, void __user *argp) ret = -EINVAL; } out: - atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); mnt_drop_write_file(file); return ret; } -- cgit v1.1