From ba7b6e62f420f5a8832bc161ab0c7ba767f65b3d Mon Sep 17 00:00:00 2001 From: David Sterba Date: Tue, 1 Jul 2014 16:21:33 +0200 Subject: btrfs: adjust statfs calculations according to raid profiles This has been discussed in thread: http://thread.gmane.org/gmane.comp.file-systems.btrfs/32528 and this patch implements this proposal: http://thread.gmane.org/gmane.comp.file-systems.btrfs/32536 Works fine for "clean" raid profiles where the raid factor correction does the right job. Otherwise it's pessimistic and may show low space although there's still some left. The df nubmers are lightly wrong in case of mixed block groups, but this is not a major usecase and can be addressed later. The RAID56 numbers are wrong almost the same way as before and will be addressed separately. CC: Hugo Mills CC: cwillu CC: Josef Bacik Signed-off-by: David Sterba Signed-off-by: Chris Mason --- fs/btrfs/super.c | 51 +++++++++++++++++++++++++++++++++++++++++++++------ 1 file changed, 45 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 8e16bca..18cdcd1 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -1672,6 +1672,21 @@ static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes) return 0; } +/* + * Calculate numbers for 'df', pessimistic in case of mixed raid profiles. + * + * If there's a redundant raid level at DATA block groups, use the respective + * multiplier to scale the sizes. + * + * Unused device space usage is based on simulating the chunk allocator + * algorithm that respects the device sizes, order of allocations and the + * 'alloc_start' value, this is a close approximation of the actual use but + * there are other factors that may change the result (like a new metadata + * chunk). + * + * FIXME: not accurate for mixed block groups, total and free/used are ok, + * available appears slightly larger. + */ static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) { struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb); @@ -1682,6 +1697,8 @@ static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) u64 total_free_data = 0; int bits = dentry->d_sb->s_blocksize_bits; __be32 *fsid = (__be32 *)fs_info->fsid; + unsigned factor = 1; + struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; int ret; /* holding chunk_muext to avoid allocating new chunks */ @@ -1689,30 +1706,52 @@ static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) rcu_read_lock(); list_for_each_entry_rcu(found, head, list) { if (found->flags & BTRFS_BLOCK_GROUP_DATA) { + int i; + total_free_data += found->disk_total - found->disk_used; total_free_data -= btrfs_account_ro_block_groups_free_space(found); + + for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) { + if (!list_empty(&found->block_groups[i])) { + switch (i) { + case BTRFS_RAID_DUP: + case BTRFS_RAID_RAID1: + case BTRFS_RAID_RAID10: + factor = 2; + } + } + } } total_used += found->disk_used; } + rcu_read_unlock(); - buf->f_namelen = BTRFS_NAME_LEN; - buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits; - buf->f_bfree = buf->f_blocks - (total_used >> bits); - buf->f_bsize = dentry->d_sb->s_blocksize; - buf->f_type = BTRFS_SUPER_MAGIC; + buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor); + buf->f_blocks >>= bits; + buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits); + + /* Account global block reserve as used, it's in logical size already */ + spin_lock(&block_rsv->lock); + buf->f_bfree -= block_rsv->size >> bits; + spin_unlock(&block_rsv->lock); + buf->f_bavail = total_free_data; ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data); if (ret) { mutex_unlock(&fs_info->chunk_mutex); return ret; } - buf->f_bavail += total_free_data; + buf->f_bavail += div_u64(total_free_data, factor); buf->f_bavail = buf->f_bavail >> bits; mutex_unlock(&fs_info->chunk_mutex); + buf->f_type = BTRFS_SUPER_MAGIC; + buf->f_bsize = dentry->d_sb->s_blocksize; + buf->f_namelen = BTRFS_NAME_LEN; + /* We treat it as constant endianness (it doesn't matter _which_) because we want the fsid to come out the same whether mounted on a big-endian or little-endian host */ -- cgit v1.1 From e339a6b097c515a31ce230d498c44ff2e89f1cf4 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 2 Jul 2014 10:54:25 -0700 Subject: Btrfs: __btrfs_mod_ref should always use no_quota Before I extended the no_quota arg to btrfs_dec/inc_ref because I didn't understand how snapshot delete was using it and assumed that we needed the quota operations there. With Mark's work this has turned out to be not the case, we _always_ need to use no_quota for btrfs_dec/inc_ref, so just drop the argument and make __btrfs_mod_ref call it's process function with no_quota set always. Thanks, Signed-off-by: Josef Bacik Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 20 ++++++++++---------- fs/btrfs/ctree.h | 4 ++-- fs/btrfs/extent-tree.c | 24 +++++++++++------------- 3 files changed, 23 insertions(+), 25 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index aeab453..44ee5d2 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -280,9 +280,9 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans, WARN_ON(btrfs_header_generation(buf) > trans->transid); if (new_root_objectid == BTRFS_TREE_RELOC_OBJECTID) - ret = btrfs_inc_ref(trans, root, cow, 1, 1); + ret = btrfs_inc_ref(trans, root, cow, 1); else - ret = btrfs_inc_ref(trans, root, cow, 0, 1); + ret = btrfs_inc_ref(trans, root, cow, 0); if (ret) return ret; @@ -1035,14 +1035,14 @@ static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans, if ((owner == root->root_key.objectid || root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) && !(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) { - ret = btrfs_inc_ref(trans, root, buf, 1, 1); + ret = btrfs_inc_ref(trans, root, buf, 1); BUG_ON(ret); /* -ENOMEM */ if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) { - ret = btrfs_dec_ref(trans, root, buf, 0, 1); + ret = btrfs_dec_ref(trans, root, buf, 0); BUG_ON(ret); /* -ENOMEM */ - ret = btrfs_inc_ref(trans, root, cow, 1, 1); + ret = btrfs_inc_ref(trans, root, cow, 1); BUG_ON(ret); /* -ENOMEM */ } new_flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF; @@ -1050,9 +1050,9 @@ static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans, if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) - ret = btrfs_inc_ref(trans, root, cow, 1, 1); + ret = btrfs_inc_ref(trans, root, cow, 1); else - ret = btrfs_inc_ref(trans, root, cow, 0, 1); + ret = btrfs_inc_ref(trans, root, cow, 0); BUG_ON(ret); /* -ENOMEM */ } if (new_flags != 0) { @@ -1069,11 +1069,11 @@ static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans, if (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF) { if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) - ret = btrfs_inc_ref(trans, root, cow, 1, 1); + ret = btrfs_inc_ref(trans, root, cow, 1); else - ret = btrfs_inc_ref(trans, root, cow, 0, 1); + ret = btrfs_inc_ref(trans, root, cow, 0); BUG_ON(ret); /* -ENOMEM */ - ret = btrfs_dec_ref(trans, root, buf, 1, 1); + ret = btrfs_dec_ref(trans, root, buf, 1); BUG_ON(ret); /* -ENOMEM */ } clean_tree_block(trans, root, buf); diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index be91397..8e29b61 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -3326,9 +3326,9 @@ int btrfs_reserve_extent(struct btrfs_root *root, u64 num_bytes, u64 min_alloc_size, u64 empty_size, u64 hint_byte, struct btrfs_key *ins, int is_data, int delalloc); int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct extent_buffer *buf, int full_backref, int no_quota); + struct extent_buffer *buf, int full_backref); int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct extent_buffer *buf, int full_backref, int no_quota); + struct extent_buffer *buf, int full_backref); int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, u64 flags, diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 813537f..591893f 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3057,7 +3057,7 @@ out: static int __btrfs_mod_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, - int full_backref, int inc, int no_quota) + int full_backref, int inc) { u64 bytenr; u64 num_bytes; @@ -3111,7 +3111,7 @@ static int __btrfs_mod_ref(struct btrfs_trans_handle *trans, key.offset -= btrfs_file_extent_offset(buf, fi); ret = process_func(trans, root, bytenr, num_bytes, parent, ref_root, key.objectid, - key.offset, no_quota); + key.offset, 1); if (ret) goto fail; } else { @@ -3119,7 +3119,7 @@ static int __btrfs_mod_ref(struct btrfs_trans_handle *trans, num_bytes = btrfs_level_size(root, level - 1); ret = process_func(trans, root, bytenr, num_bytes, parent, ref_root, level - 1, 0, - no_quota); + 1); if (ret) goto fail; } @@ -3130,15 +3130,15 @@ fail: } int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct extent_buffer *buf, int full_backref, int no_quota) + struct extent_buffer *buf, int full_backref) { - return __btrfs_mod_ref(trans, root, buf, full_backref, 1, no_quota); + return __btrfs_mod_ref(trans, root, buf, full_backref, 1); } int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct extent_buffer *buf, int full_backref, int no_quota) + struct extent_buffer *buf, int full_backref) { - return __btrfs_mod_ref(trans, root, buf, full_backref, 0, no_quota); + return __btrfs_mod_ref(trans, root, buf, full_backref, 0); } static int write_one_cache_group(struct btrfs_trans_handle *trans, @@ -7532,9 +7532,9 @@ static noinline int walk_down_proc(struct btrfs_trans_handle *trans, /* wc->stage == UPDATE_BACKREF */ if (!(wc->flags[level] & flag)) { BUG_ON(!path->locks[level]); - ret = btrfs_inc_ref(trans, root, eb, 1, wc->for_reloc); + ret = btrfs_inc_ref(trans, root, eb, 1); BUG_ON(ret); /* -ENOMEM */ - ret = btrfs_dec_ref(trans, root, eb, 0, wc->for_reloc); + ret = btrfs_dec_ref(trans, root, eb, 0); BUG_ON(ret); /* -ENOMEM */ ret = btrfs_set_disk_extent_flags(trans, root, eb->start, eb->len, flag, @@ -7769,11 +7769,9 @@ static noinline int walk_up_proc(struct btrfs_trans_handle *trans, if (wc->refs[level] == 1) { if (level == 0) { if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) - ret = btrfs_dec_ref(trans, root, eb, 1, - wc->for_reloc); + ret = btrfs_dec_ref(trans, root, eb, 1); else - ret = btrfs_dec_ref(trans, root, eb, 0, - wc->for_reloc); + ret = btrfs_dec_ref(trans, root, eb, 0); BUG_ON(ret); /* -ENOMEM */ } /* make block locked assertion in clean_tree_block happy */ -- cgit v1.1 From 6f7ff6d7832c6be13e8c95598884dbc40ad69fb7 Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Wed, 2 Jul 2014 20:07:54 +0100 Subject: Btrfs: read lock extent buffer while walking backrefs Before processing the extent buffer, acquire a read lock on it, so that we're safe against concurrent updates on the extent buffer. Signed-off-by: Filipe Manana Signed-off-by: Chris Mason --- fs/btrfs/backref.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'fs') diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c index e25564b..a1efd39 100644 --- a/fs/btrfs/backref.c +++ b/fs/btrfs/backref.c @@ -1001,8 +1001,11 @@ again: ret = -EIO; goto out; } + btrfs_tree_read_lock(eb); + btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); ret = find_extent_in_eb(eb, bytenr, *extent_item_pos, &eie); + btrfs_tree_read_unlock_blocking(eb); free_extent_buffer(eb); if (ret < 0) goto out; -- cgit v1.1 From 1152651a081720ef6a8c76bb7da676e8c900ac30 Mon Sep 17 00:00:00 2001 From: Mark Fasheh Date: Thu, 17 Jul 2014 12:39:01 -0700 Subject: btrfs: qgroup: account shared subtrees during snapshot delete During its tree walk, btrfs_drop_snapshot() will skip any shared subtrees it encounters. This is incorrect when we have qgroups turned on as those subtrees need to have their contents accounted. In particular, the case we're concerned with is when removing our snapshot root leaves the subtree with only one root reference. In those cases we need to find the last remaining root and add each extent in the subtree to the corresponding qgroup exclusive counts. This patch implements the shared subtree walk and a new qgroup operation, BTRFS_QGROUP_OPER_SUB_SUBTREE. When an operation of this type is encountered during qgroup accounting, we search for any root references to that extent and in the case that we find only one reference left, we go ahead and do the math on it's exclusive counts. Signed-off-by: Mark Fasheh Reviewed-by: Josef Bacik Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 261 +++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/qgroup.c | 164 +++++++++++++++++++++++++++++++ fs/btrfs/qgroup.h | 1 + 3 files changed, 426 insertions(+) (limited to 'fs') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 591893f..102ed31 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -7478,6 +7478,220 @@ reada: wc->reada_slot = slot; } +static int account_leaf_items(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct extent_buffer *eb) +{ + int nr = btrfs_header_nritems(eb); + int i, extent_type, ret; + struct btrfs_key key; + struct btrfs_file_extent_item *fi; + u64 bytenr, num_bytes; + + for (i = 0; i < nr; i++) { + btrfs_item_key_to_cpu(eb, &key, i); + + if (key.type != BTRFS_EXTENT_DATA_KEY) + continue; + + fi = btrfs_item_ptr(eb, i, struct btrfs_file_extent_item); + /* filter out non qgroup-accountable extents */ + extent_type = btrfs_file_extent_type(eb, fi); + + if (extent_type == BTRFS_FILE_EXTENT_INLINE) + continue; + + bytenr = btrfs_file_extent_disk_bytenr(eb, fi); + if (!bytenr) + continue; + + num_bytes = btrfs_file_extent_disk_num_bytes(eb, fi); + + ret = btrfs_qgroup_record_ref(trans, root->fs_info, + root->objectid, + bytenr, num_bytes, + BTRFS_QGROUP_OPER_SUB_SUBTREE, 0); + if (ret) + return ret; + } + return 0; +} + +/* + * Walk up the tree from the bottom, freeing leaves and any interior + * nodes which have had all slots visited. If a node (leaf or + * interior) is freed, the node above it will have it's slot + * incremented. The root node will never be freed. + * + * At the end of this function, we should have a path which has all + * slots incremented to the next position for a search. If we need to + * read a new node it will be NULL and the node above it will have the + * correct slot selected for a later read. + * + * If we increment the root nodes slot counter past the number of + * elements, 1 is returned to signal completion of the search. + */ +static int adjust_slots_upwards(struct btrfs_root *root, + struct btrfs_path *path, int root_level) +{ + int level = 0; + int nr, slot; + struct extent_buffer *eb; + + if (root_level == 0) + return 1; + + while (level <= root_level) { + eb = path->nodes[level]; + nr = btrfs_header_nritems(eb); + path->slots[level]++; + slot = path->slots[level]; + if (slot >= nr || level == 0) { + /* + * Don't free the root - we will detect this + * condition after our loop and return a + * positive value for caller to stop walking the tree. + */ + if (level != root_level) { + btrfs_tree_unlock_rw(eb, path->locks[level]); + path->locks[level] = 0; + + free_extent_buffer(eb); + path->nodes[level] = NULL; + path->slots[level] = 0; + } + } else { + /* + * We have a valid slot to walk back down + * from. Stop here so caller can process these + * new nodes. + */ + break; + } + + level++; + } + + eb = path->nodes[root_level]; + if (path->slots[root_level] >= btrfs_header_nritems(eb)) + return 1; + + return 0; +} + +/* + * root_eb is the subtree root and is locked before this function is called. + */ +static int account_shared_subtree(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct extent_buffer *root_eb, + u64 root_gen, + int root_level) +{ + int ret = 0; + int level; + struct extent_buffer *eb = root_eb; + struct btrfs_path *path = NULL; + + BUG_ON(root_level < 0 || root_level > BTRFS_MAX_LEVEL); + BUG_ON(root_eb == NULL); + + if (!root->fs_info->quota_enabled) + return 0; + + if (!extent_buffer_uptodate(root_eb)) { + ret = btrfs_read_buffer(root_eb, root_gen); + if (ret) + goto out; + } + + if (root_level == 0) { + ret = account_leaf_items(trans, root, root_eb); + goto out; + } + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + /* + * Walk down the tree. Missing extent blocks are filled in as + * we go. Metadata is accounted every time we read a new + * extent block. + * + * When we reach a leaf, we account for file extent items in it, + * walk back up the tree (adjusting slot pointers as we go) + * and restart the search process. + */ + extent_buffer_get(root_eb); /* For path */ + path->nodes[root_level] = root_eb; + path->slots[root_level] = 0; + path->locks[root_level] = 0; /* so release_path doesn't try to unlock */ +walk_down: + level = root_level; + while (level >= 0) { + if (path->nodes[level] == NULL) { + int child_bsize = root->nodesize; + int parent_slot; + u64 child_gen; + u64 child_bytenr; + + /* We need to get child blockptr/gen from + * parent before we can read it. */ + eb = path->nodes[level + 1]; + parent_slot = path->slots[level + 1]; + child_bytenr = btrfs_node_blockptr(eb, parent_slot); + child_gen = btrfs_node_ptr_generation(eb, parent_slot); + + eb = read_tree_block(root, child_bytenr, child_bsize, + child_gen); + if (!eb || !extent_buffer_uptodate(eb)) { + ret = -EIO; + goto out; + } + + path->nodes[level] = eb; + path->slots[level] = 0; + + btrfs_tree_read_lock(eb); + btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); + path->locks[level] = BTRFS_READ_LOCK_BLOCKING; + + ret = btrfs_qgroup_record_ref(trans, root->fs_info, + root->objectid, + child_bytenr, + child_bsize, + BTRFS_QGROUP_OPER_SUB_SUBTREE, + 0); + if (ret) + goto out; + + } + + if (level == 0) { + ret = account_leaf_items(trans, root, path->nodes[level]); + if (ret) + goto out; + + /* Nonzero return here means we completed our search */ + ret = adjust_slots_upwards(root, path, root_level); + if (ret) + break; + + /* Restart search with new slots */ + goto walk_down; + } + + level--; + } + + ret = 0; +out: + btrfs_free_path(path); + + return ret; +} + /* * helper to process tree block while walking down the tree. * @@ -7581,6 +7795,7 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans, int level = wc->level; int reada = 0; int ret = 0; + bool need_account = false; generation = btrfs_node_ptr_generation(path->nodes[level], path->slots[level]); @@ -7626,6 +7841,7 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans, if (wc->stage == DROP_REFERENCE) { if (wc->refs[level - 1] > 1) { + need_account = true; if (level == 1 && (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF)) goto skip; @@ -7689,6 +7905,16 @@ skip: parent = 0; } + if (need_account) { + ret = account_shared_subtree(trans, root, next, + generation, level - 1); + if (ret) { + printk_ratelimited(KERN_ERR "BTRFS: %s Error " + "%d accounting shared subtree. Quota " + "is out of sync, rescan required.\n", + root->fs_info->sb->s_id, ret); + } + } ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent, root->root_key.objectid, level - 1, 0, 0); BUG_ON(ret); /* -ENOMEM */ @@ -7773,6 +7999,13 @@ static noinline int walk_up_proc(struct btrfs_trans_handle *trans, else ret = btrfs_dec_ref(trans, root, eb, 0); BUG_ON(ret); /* -ENOMEM */ + ret = account_leaf_items(trans, root, eb); + if (ret) { + printk_ratelimited(KERN_ERR "BTRFS: %s Error " + "%d accounting leaf items. Quota " + "is out of sync, rescan required.\n", + root->fs_info->sb->s_id, ret); + } } /* make block locked assertion in clean_tree_block happy */ if (!path->locks[level] && @@ -7898,6 +8131,8 @@ int btrfs_drop_snapshot(struct btrfs_root *root, int level; bool root_dropped = false; + btrfs_debug(root->fs_info, "Drop subvolume %llu", root->objectid); + path = btrfs_alloc_path(); if (!path) { err = -ENOMEM; @@ -8023,6 +8258,24 @@ int btrfs_drop_snapshot(struct btrfs_root *root, goto out_end_trans; } + /* + * Qgroup update accounting is run from + * delayed ref handling. This usually works + * out because delayed refs are normally the + * only way qgroup updates are added. However, + * we may have added updates during our tree + * walk so run qgroups here to make sure we + * don't lose any updates. + */ + ret = btrfs_delayed_qgroup_accounting(trans, + root->fs_info); + if (ret) + printk_ratelimited(KERN_ERR "BTRFS: Failure %d " + "running qgroup updates " + "during snapshot delete. " + "Quota is out of sync, " + "rescan required.\n", ret); + btrfs_end_transaction_throttle(trans, tree_root); if (!for_reloc && btrfs_need_cleaner_sleep(root)) { pr_debug("BTRFS: drop snapshot early exit\n"); @@ -8076,6 +8329,14 @@ int btrfs_drop_snapshot(struct btrfs_root *root, } root_dropped = true; out_end_trans: + ret = btrfs_delayed_qgroup_accounting(trans, tree_root->fs_info); + if (ret) + printk_ratelimited(KERN_ERR "BTRFS: Failure %d " + "running qgroup updates " + "during snapshot delete. " + "Quota is out of sync, " + "rescan required.\n", ret); + btrfs_end_transaction_throttle(trans, tree_root); out_free: kfree(wc); diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c index 98cb6b2..65b62c4 100644 --- a/fs/btrfs/qgroup.c +++ b/fs/btrfs/qgroup.c @@ -1201,6 +1201,50 @@ out: mutex_unlock(&fs_info->qgroup_ioctl_lock); return ret; } + +static int comp_oper_exist(struct btrfs_qgroup_operation *oper1, + struct btrfs_qgroup_operation *oper2) +{ + /* + * Ignore seq and type here, we're looking for any operation + * at all related to this extent on that root. + */ + if (oper1->bytenr < oper2->bytenr) + return -1; + if (oper1->bytenr > oper2->bytenr) + return 1; + if (oper1->ref_root < oper2->ref_root) + return -1; + if (oper1->ref_root > oper2->ref_root) + return 1; + return 0; +} + +static int qgroup_oper_exists(struct btrfs_fs_info *fs_info, + struct btrfs_qgroup_operation *oper) +{ + struct rb_node *n; + struct btrfs_qgroup_operation *cur; + int cmp; + + spin_lock(&fs_info->qgroup_op_lock); + n = fs_info->qgroup_op_tree.rb_node; + while (n) { + cur = rb_entry(n, struct btrfs_qgroup_operation, n); + cmp = comp_oper_exist(cur, oper); + if (cmp < 0) { + n = n->rb_right; + } else if (cmp) { + n = n->rb_left; + } else { + spin_unlock(&fs_info->qgroup_op_lock); + return -EEXIST; + } + } + spin_unlock(&fs_info->qgroup_op_lock); + return 0; +} + static int comp_oper(struct btrfs_qgroup_operation *oper1, struct btrfs_qgroup_operation *oper2) { @@ -1290,6 +1334,23 @@ int btrfs_qgroup_record_ref(struct btrfs_trans_handle *trans, oper->seq = atomic_inc_return(&fs_info->qgroup_op_seq); INIT_LIST_HEAD(&oper->elem.list); oper->elem.seq = 0; + + if (type == BTRFS_QGROUP_OPER_SUB_SUBTREE) { + /* + * If any operation for this bytenr/ref_root combo + * exists, then we know it's not exclusively owned and + * shouldn't be queued up. + * + * This also catches the case where we have a cloned + * extent that gets queued up multiple times during + * drop snapshot. + */ + if (qgroup_oper_exists(fs_info, oper)) { + kfree(oper); + return 0; + } + } + ret = insert_qgroup_oper(fs_info, oper); if (ret) { /* Shouldn't happen so have an assert for developers */ @@ -1884,6 +1945,106 @@ out: } /* + * Process a reference to a shared subtree. This type of operation is + * queued during snapshot removal when we encounter extents which are + * shared between more than one root. + */ +static int qgroup_subtree_accounting(struct btrfs_trans_handle *trans, + struct btrfs_fs_info *fs_info, + struct btrfs_qgroup_operation *oper) +{ + struct ulist *roots = NULL; + struct ulist_node *unode; + struct ulist_iterator uiter; + struct btrfs_qgroup_list *glist; + struct ulist *parents; + int ret = 0; + struct btrfs_qgroup *qg; + u64 root_obj = 0; + struct seq_list elem = {}; + + parents = ulist_alloc(GFP_NOFS); + if (!parents) + return -ENOMEM; + + btrfs_get_tree_mod_seq(fs_info, &elem); + ret = btrfs_find_all_roots(trans, fs_info, oper->bytenr, + elem.seq, &roots); + btrfs_put_tree_mod_seq(fs_info, &elem); + if (ret < 0) + return ret; + + if (roots->nnodes != 1) + goto out; + + ULIST_ITER_INIT(&uiter); + unode = ulist_next(roots, &uiter); /* Only want 1 so no need to loop */ + /* + * If we find our ref root then that means all refs + * this extent has to the root have not yet been + * deleted. In that case, we do nothing and let the + * last ref for this bytenr drive our update. + * + * This can happen for example if an extent is + * referenced multiple times in a snapshot (clone, + * etc). If we are in the middle of snapshot removal, + * queued updates for such an extent will find the + * root if we have not yet finished removing the + * snapshot. + */ + if (unode->val == oper->ref_root) + goto out; + + root_obj = unode->val; + BUG_ON(!root_obj); + + spin_lock(&fs_info->qgroup_lock); + qg = find_qgroup_rb(fs_info, root_obj); + if (!qg) + goto out_unlock; + + qg->excl += oper->num_bytes; + qg->excl_cmpr += oper->num_bytes; + qgroup_dirty(fs_info, qg); + + /* + * Adjust counts for parent groups. First we find all + * parents, then in the 2nd loop we do the adjustment + * while adding parents of the parents to our ulist. + */ + list_for_each_entry(glist, &qg->groups, next_group) { + ret = ulist_add(parents, glist->group->qgroupid, + ptr_to_u64(glist->group), GFP_ATOMIC); + if (ret < 0) + goto out_unlock; + } + + ULIST_ITER_INIT(&uiter); + while ((unode = ulist_next(parents, &uiter))) { + qg = u64_to_ptr(unode->aux); + qg->excl += oper->num_bytes; + qg->excl_cmpr += oper->num_bytes; + qgroup_dirty(fs_info, qg); + + /* Add any parents of the parents */ + list_for_each_entry(glist, &qg->groups, next_group) { + ret = ulist_add(parents, glist->group->qgroupid, + ptr_to_u64(glist->group), GFP_ATOMIC); + if (ret < 0) + goto out_unlock; + } + } + +out_unlock: + spin_unlock(&fs_info->qgroup_lock); + +out: + ulist_free(roots); + ulist_free(parents); + return ret; +} + +/* * btrfs_qgroup_account_ref is called for every ref that is added to or deleted * from the fs. First, all roots referencing the extent are searched, and * then the space is accounted accordingly to the different roots. The @@ -1920,6 +2081,9 @@ static int btrfs_qgroup_account(struct btrfs_trans_handle *trans, case BTRFS_QGROUP_OPER_SUB_SHARED: ret = qgroup_shared_accounting(trans, fs_info, oper); break; + case BTRFS_QGROUP_OPER_SUB_SUBTREE: + ret = qgroup_subtree_accounting(trans, fs_info, oper); + break; default: ASSERT(0); } diff --git a/fs/btrfs/qgroup.h b/fs/btrfs/qgroup.h index 5952ff1..18cc68c 100644 --- a/fs/btrfs/qgroup.h +++ b/fs/btrfs/qgroup.h @@ -44,6 +44,7 @@ enum btrfs_qgroup_operation_type { BTRFS_QGROUP_OPER_ADD_SHARED, BTRFS_QGROUP_OPER_SUB_EXCL, BTRFS_QGROUP_OPER_SUB_SHARED, + BTRFS_QGROUP_OPER_SUB_SUBTREE, }; struct btrfs_qgroup_operation { -- cgit v1.1 From f90e579c2b391979630b3343de0be65ab1b478ce Mon Sep 17 00:00:00 2001 From: Mark Fasheh Date: Thu, 17 Jul 2014 12:39:04 -0700 Subject: btrfs: correctly handle return from ulist_add ulist_add() can return '1' on sucess, which qgroup_subtree_accounting() doesn't take into account. As a result, that value can be bubbled up to callers, causing an error to be printed. Fix this by only returning the value of ulist_add() when it indicates an error. Signed-off-by: Mark Fasheh Signed-off-by: Chris Mason --- fs/btrfs/qgroup.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c index 65b62c4..b497498 100644 --- a/fs/btrfs/qgroup.c +++ b/fs/btrfs/qgroup.c @@ -1959,6 +1959,7 @@ static int qgroup_subtree_accounting(struct btrfs_trans_handle *trans, struct btrfs_qgroup_list *glist; struct ulist *parents; int ret = 0; + int err; struct btrfs_qgroup *qg; u64 root_obj = 0; struct seq_list elem = {}; @@ -2013,10 +2014,12 @@ static int qgroup_subtree_accounting(struct btrfs_trans_handle *trans, * while adding parents of the parents to our ulist. */ list_for_each_entry(glist, &qg->groups, next_group) { - ret = ulist_add(parents, glist->group->qgroupid, + err = ulist_add(parents, glist->group->qgroupid, ptr_to_u64(glist->group), GFP_ATOMIC); - if (ret < 0) + if (err < 0) { + ret = err; goto out_unlock; + } } ULIST_ITER_INIT(&uiter); @@ -2028,10 +2031,12 @@ static int qgroup_subtree_accounting(struct btrfs_trans_handle *trans, /* Add any parents of the parents */ list_for_each_entry(glist, &qg->groups, next_group) { - ret = ulist_add(parents, glist->group->qgroupid, + err = ulist_add(parents, glist->group->qgroupid, ptr_to_u64(glist->group), GFP_ATOMIC); - if (ret < 0) + if (err < 0) { + ret = err; goto out_unlock; + } } } -- cgit v1.1 From ce62003f690dff38d3164a632ec69efa15c32cbf Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Thu, 24 Jul 2014 22:48:05 +0800 Subject: Btrfs: fix compressed write corruption on enospc When failing to allocate space for the whole compressed extent, we'll fallback to uncompressed IO, but we've forgotten to redirty the pages which belong to this compressed extent, and these 'clean' pages will simply skip 'submit' part and go to endio directly, at last we got data corruption as we write nothing. Signed-off-by: Liu Bo Tested-By: Martin Steigerwald Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'fs') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3668048..8ea7610 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -709,6 +709,18 @@ retry: unlock_extent(io_tree, async_extent->start, async_extent->start + async_extent->ram_size - 1); + + /* + * we need to redirty the pages if we decide to + * fallback to uncompressed IO, otherwise we + * will not submit these pages down to lower + * layers. + */ + extent_range_redirty_for_io(inode, + async_extent->start, + async_extent->start + + async_extent->ram_size - 1); + goto retry; } goto out_free; -- cgit v1.1 From 4eb1f66dce6c4dc28dd90a7ffbe6b2b1cb08aa4e Mon Sep 17 00:00:00 2001 From: Takashi Iwai Date: Mon, 28 Jul 2014 10:57:04 +0200 Subject: Btrfs: Fix memory corruption by ulist_add_merge() on 32bit arch We've got bug reports that btrfs crashes when quota is enabled on 32bit kernel, typically with the Oops like below: BUG: unable to handle kernel NULL pointer dereference at 00000004 IP: [] find_parent_nodes+0x360/0x1380 [btrfs] *pde = 00000000 Oops: 0000 [#1] SMP CPU: 0 PID: 151 Comm: kworker/u8:2 Tainted: G S W 3.15.2-1.gd43d97e-default #1 Workqueue: btrfs-qgroup-rescan normal_work_helper [btrfs] task: f1478130 ti: f147c000 task.ti: f147c000 EIP: 0060:[] EFLAGS: 00010213 CPU: 0 EIP is at find_parent_nodes+0x360/0x1380 [btrfs] EAX: f147dda8 EBX: f147ddb0 ECX: 00000011 EDX: 00000000 ESI: 00000000 EDI: f147dda4 EBP: f147ddf8 ESP: f147dd38 DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068 CR0: 8005003b CR2: 00000004 CR3: 00bf3000 CR4: 00000690 Stack: 00000000 00000000 f147dda4 00000050 00000001 00000000 00000001 00000050 00000001 00000000 d3059000 00000001 00000022 000000a8 00000000 00000000 00000000 000000a1 00000000 00000000 00000001 00000000 00000000 11800000 Call Trace: [] __btrfs_find_all_roots+0x9d/0xf0 [btrfs] [] btrfs_qgroup_rescan_worker+0x401/0x760 [btrfs] [] normal_work_helper+0xc8/0x270 [btrfs] [] process_one_work+0x11b/0x390 [] worker_thread+0x101/0x340 [] kthread+0x9b/0xb0 [] ret_from_kernel_thread+0x21/0x30 [] kthread_create_on_node+0x110/0x110 This indicates a NULL corruption in prefs_delayed list. The further investigation and bisection pointed that the call of ulist_add_merge() results in the corruption. ulist_add_merge() takes u64 as aux and writes a 64bit value into old_aux. The callers of this function in backref.c, however, pass a pointer of a pointer to old_aux. That is, the function overwrites 64bit value on 32bit pointer. This caused a NULL in the adjacent variable, in this case, prefs_delayed. Here is a quick attempt to band-aid over this: a new function, ulist_add_merge_ptr() is introduced to pass/store properly a pointer value instead of u64. There are still ugly void ** cast remaining in the callers because void ** cannot be taken implicitly. But, it's safer than explicit cast to u64, anyway. Bugzilla: https://bugzilla.novell.com/show_bug.cgi?id=887046 Cc: [v3.11+] Signed-off-by: Takashi Iwai Signed-off-by: Chris Mason --- fs/btrfs/backref.c | 11 +++++------ fs/btrfs/ulist.h | 15 +++++++++++++++ 2 files changed, 20 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c index a1efd39..54a201d 100644 --- a/fs/btrfs/backref.c +++ b/fs/btrfs/backref.c @@ -276,9 +276,8 @@ static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path, } if (ret > 0) goto next; - ret = ulist_add_merge(parents, eb->start, - (uintptr_t)eie, - (u64 *)&old, GFP_NOFS); + ret = ulist_add_merge_ptr(parents, eb->start, + eie, (void **)&old, GFP_NOFS); if (ret < 0) break; if (!ret && extent_item_pos) { @@ -1011,9 +1010,9 @@ again: goto out; ref->inode_list = eie; } - ret = ulist_add_merge(refs, ref->parent, - (uintptr_t)ref->inode_list, - (u64 *)&eie, GFP_NOFS); + ret = ulist_add_merge_ptr(refs, ref->parent, + ref->inode_list, + (void **)&eie, GFP_NOFS); if (ret < 0) goto out; if (!ret && extent_item_pos) { diff --git a/fs/btrfs/ulist.h b/fs/btrfs/ulist.h index 7f78cbf..4c29db6 100644 --- a/fs/btrfs/ulist.h +++ b/fs/btrfs/ulist.h @@ -57,6 +57,21 @@ void ulist_free(struct ulist *ulist); int ulist_add(struct ulist *ulist, u64 val, u64 aux, gfp_t gfp_mask); int ulist_add_merge(struct ulist *ulist, u64 val, u64 aux, u64 *old_aux, gfp_t gfp_mask); + +/* just like ulist_add_merge() but take a pointer for the aux data */ +static inline int ulist_add_merge_ptr(struct ulist *ulist, u64 val, void *aux, + void **old_aux, gfp_t gfp_mask) +{ +#if BITS_PER_LONG == 32 + u64 old64 = (uintptr_t)*old_aux; + int ret = ulist_add_merge(ulist, val, (uintptr_t)aux, &old64, gfp_mask); + *old_aux = (void *)((uintptr_t)old64); + return ret; +#else + return ulist_add_merge(ulist, val, (u64)aux, (u64 *)old_aux, gfp_mask); +#endif +} + struct ulist_node *ulist_next(struct ulist *ulist, struct ulist_iterator *uiter); -- cgit v1.1 From 27b9a8122ff71a8cadfbffb9c4f0694300464f3b Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Sat, 9 Aug 2014 21:22:27 +0100 Subject: Btrfs: fix csum tree corruption, duplicate and outdated checksums Under rare circumstances we can end up leaving 2 versions of a checksum for the same file extent range. The reason for this is that after calling btrfs_next_leaf we process slot 0 of the leaf it returns, instead of processing the slot set in path->slots[0]. Most of the time (by far) path->slots[0] is 0, but after btrfs_next_leaf() releases the path and before it searches for the next leaf, another task might cause a split of the next leaf, which migrates some of its keys to the leaf we were processing before calling btrfs_next_leaf(). In this case btrfs_next_leaf() returns again the same leaf but with path->slots[0] having a slot number corresponding to the first new key it got, that is, a slot number that didn't exist before calling btrfs_next_leaf(), as the leaf now has more keys than it had before. So we must really process the returned leaf starting at path->slots[0] always, as it isn't always 0, and the key at slot 0 can have an offset much lower than our search offset/bytenr. For example, consider the following scenario, where we have: sums->bytenr: 40157184, sums->len: 16384, sums end: 40173568 four 4kb file data blocks with offsets 40157184, 40161280, 40165376, 40169472 Leaf N: slot = 0 slot = btrfs_header_nritems() - 1 |-------------------------------------------------------------------| | [(CSUM CSUM 39239680), size 8] ... [(CSUM CSUM 40116224), size 4] | |-------------------------------------------------------------------| Leaf N + 1: slot = 0 slot = btrfs_header_nritems() - 1 |--------------------------------------------------------------------| | [(CSUM CSUM 40161280), size 32] ... [((CSUM CSUM 40615936), size 8 | |--------------------------------------------------------------------| Because we are at the last slot of leaf N, we call btrfs_next_leaf() to find the next highest key, which releases the current path and then searches for that next key. However after releasing the path and before finding that next key, the item at slot 0 of leaf N + 1 gets moved to leaf N, due to a call to ctree.c:push_leaf_left() (via ctree.c:split_leaf()), and therefore btrfs_next_leaf() will returns us a path again with leaf N but with the slot pointing to its new last key (CSUM CSUM 40161280). This new version of leaf N is then: slot = 0 slot = btrfs_header_nritems() - 2 slot = btrfs_header_nritems() - 1 |----------------------------------------------------------------------------------------------------| | [(CSUM CSUM 39239680), size 8] ... [(CSUM CSUM 40116224), size 4] [(CSUM CSUM 40161280), size 32] | |----------------------------------------------------------------------------------------------------| And incorrecly using slot 0, makes us set next_offset to 39239680 and we jump into the "insert:" label, which will set tmp to: tmp = min((sums->len - total_bytes) >> blocksize_bits, (next_offset - file_key.offset) >> blocksize_bits) = min((16384 - 0) >> 12, (39239680 - 40157184) >> 12) = min(4, (u64)-917504 = 18446744073708634112 >> 12) = 4 and ins_size = csum_size * tmp = 4 * 4 = 16 bytes. In other words, we insert a new csum item in the tree with key (CSUM_OBJECTID CSUM_KEY 40157184 = sums->bytenr) that contains the checksums for all the data (4 blocks of 4096 bytes each = sums->len). Which is wrong, because the item with key (CSUM CSUM 40161280) (the one that was moved from leaf N + 1 to the end of leaf N) contains the old checksums of the last 12288 bytes of our data and won't get those old checksums removed. So this leaves us 2 different checksums for 3 4kb blocks of data in the tree, and breaks the logical rule: Key_N+1.offset >= Key_N.offset + length_of_data_its_checksums_cover An obvious bad effect of this is that a subsequent csum tree lookup to get the checksum of any of the blocks with logical offset of 40161280, 40165376 or 40169472 (the last 3 4kb blocks of file data), will get the old checksums. Cc: stable@vger.kernel.org Signed-off-by: Filipe Manana Signed-off-by: Chris Mason --- fs/btrfs/file-item.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index f46cfe4..54c84da 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -756,7 +756,7 @@ again: found_next = 1; if (ret != 0) goto insert; - slot = 0; + slot = path->slots[0]; } btrfs_item_key_to_cpu(path->nodes[0], &found_key, slot); if (found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID || -- cgit v1.1 From 8d875f95da43c6a8f18f77869f2ef26e9594fecc Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 12 Aug 2014 10:47:42 -0700 Subject: btrfs: disable strict file flushes for renames and truncates Truncates and renames are often used to replace old versions of a file with new versions. Applications often expect this to be an atomic replacement, even if they haven't done anything to make sure the new version is fully on disk. Btrfs has strict flushing in place to make sure that renaming over an old file with a new file will fully flush out the new file before allowing the transaction commit with the rename to complete. This ordering means the commit code needs to be able to lock file pages, and there are a few paths in the filesystem where we will try to end a transaction with the page lock held. It's rare, but these things can deadlock. This patch removes the ordered flushes and switches to a best effort filemap_flush like ext4 uses. It's not perfect, but it should fix the deadlocks. Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 6 --- fs/btrfs/disk-io.c | 32 ------------- fs/btrfs/file.c | 26 +--------- fs/btrfs/inode.c | 47 ++---------------- fs/btrfs/ordered-data.c | 123 ------------------------------------------------ fs/btrfs/ordered-data.h | 5 -- fs/btrfs/transaction.c | 33 +------------ fs/btrfs/transaction.h | 1 - 8 files changed, 6 insertions(+), 267 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 4794923..43527fd 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -84,12 +84,6 @@ struct btrfs_inode { */ struct list_head delalloc_inodes; - /* - * list for tracking inodes that must be sent to disk before a - * rename or truncate commit - */ - struct list_head ordered_operations; - /* node for the red-black tree that links inodes in subvolume root */ struct rb_node rb_node; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 08e65e9..d0ed9e6 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -60,8 +60,6 @@ static void end_workqueue_fn(struct btrfs_work *work); static void free_fs_root(struct btrfs_root *root); static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, int read_only); -static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, - struct btrfs_root *root); static void btrfs_destroy_ordered_extents(struct btrfs_root *root); static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, struct btrfs_root *root); @@ -3829,34 +3827,6 @@ static void btrfs_error_commit_super(struct btrfs_root *root) btrfs_cleanup_transaction(root); } -static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, - struct btrfs_root *root) -{ - struct btrfs_inode *btrfs_inode; - struct list_head splice; - - INIT_LIST_HEAD(&splice); - - mutex_lock(&root->fs_info->ordered_operations_mutex); - spin_lock(&root->fs_info->ordered_root_lock); - - list_splice_init(&t->ordered_operations, &splice); - while (!list_empty(&splice)) { - btrfs_inode = list_entry(splice.next, struct btrfs_inode, - ordered_operations); - - list_del_init(&btrfs_inode->ordered_operations); - spin_unlock(&root->fs_info->ordered_root_lock); - - btrfs_invalidate_inodes(btrfs_inode->root); - - spin_lock(&root->fs_info->ordered_root_lock); - } - - spin_unlock(&root->fs_info->ordered_root_lock); - mutex_unlock(&root->fs_info->ordered_operations_mutex); -} - static void btrfs_destroy_ordered_extents(struct btrfs_root *root) { struct btrfs_ordered_extent *ordered; @@ -4093,8 +4063,6 @@ again: void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, struct btrfs_root *root) { - btrfs_destroy_ordered_operations(cur_trans, root); - btrfs_destroy_delayed_refs(cur_trans, root); cur_trans->state = TRANS_STATE_COMMIT_START; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 1f2b99c..d3afac2 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1838,33 +1838,9 @@ out: int btrfs_release_file(struct inode *inode, struct file *filp) { - /* - * ordered_data_close is set by settattr when we are about to truncate - * a file from a non-zero size to a zero size. This tries to - * flush down new bytes that may have been written if the - * application were using truncate to replace a file in place. - */ - if (test_and_clear_bit(BTRFS_INODE_ORDERED_DATA_CLOSE, - &BTRFS_I(inode)->runtime_flags)) { - struct btrfs_trans_handle *trans; - struct btrfs_root *root = BTRFS_I(inode)->root; - - /* - * We need to block on a committing transaction to keep us from - * throwing a ordered operation on to the list and causing - * something like sync to deadlock trying to flush out this - * inode. - */ - trans = btrfs_start_transaction(root, 0); - if (IS_ERR(trans)) - return PTR_ERR(trans); - btrfs_add_ordered_operation(trans, BTRFS_I(inode)->root, inode); - btrfs_end_transaction(trans, root); - if (inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT) - filemap_flush(inode->i_mapping); - } if (filp->private_data) btrfs_ioctl_trans_end(filp); + filemap_flush(inode->i_mapping); return 0; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 8ea7610..7309832 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -7951,27 +7951,6 @@ static int btrfs_truncate(struct inode *inode) BUG_ON(ret); /* - * setattr is responsible for setting the ordered_data_close flag, - * but that is only tested during the last file release. That - * could happen well after the next commit, leaving a great big - * window where new writes may get lost if someone chooses to write - * to this file after truncating to zero - * - * The inode doesn't have any dirty data here, and so if we commit - * this is a noop. If someone immediately starts writing to the inode - * it is very likely we'll catch some of their writes in this - * transaction, and the commit will find this file on the ordered - * data list with good things to send down. - * - * This is a best effort solution, there is still a window where - * using truncate to replace the contents of the file will - * end up with a zero length file after a crash. - */ - if (inode->i_size == 0 && test_bit(BTRFS_INODE_ORDERED_DATA_CLOSE, - &BTRFS_I(inode)->runtime_flags)) - btrfs_add_ordered_operation(trans, root, inode); - - /* * So if we truncate and then write and fsync we normally would just * write the extents that changed, which is a problem if we need to * first truncate that entire inode. So set this flag so we write out @@ -8118,7 +8097,6 @@ struct inode *btrfs_alloc_inode(struct super_block *sb) mutex_init(&ei->delalloc_mutex); btrfs_ordered_inode_tree_init(&ei->ordered_tree); INIT_LIST_HEAD(&ei->delalloc_inodes); - INIT_LIST_HEAD(&ei->ordered_operations); RB_CLEAR_NODE(&ei->rb_node); return inode; @@ -8158,17 +8136,6 @@ void btrfs_destroy_inode(struct inode *inode) if (!root) goto free; - /* - * Make sure we're properly removed from the ordered operation - * lists. - */ - smp_mb(); - if (!list_empty(&BTRFS_I(inode)->ordered_operations)) { - spin_lock(&root->fs_info->ordered_root_lock); - list_del_init(&BTRFS_I(inode)->ordered_operations); - spin_unlock(&root->fs_info->ordered_root_lock); - } - if (test_bit(BTRFS_INODE_HAS_ORPHAN_ITEM, &BTRFS_I(inode)->runtime_flags)) { btrfs_info(root->fs_info, "inode %llu still on the orphan list", @@ -8350,12 +8317,10 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry, ret = 0; /* - * we're using rename to replace one file with another. - * and the replacement file is large. Start IO on it now so - * we don't add too much work to the end of the transaction + * we're using rename to replace one file with another. Start IO on it + * now so we don't add too much work to the end of the transaction */ - if (new_inode && S_ISREG(old_inode->i_mode) && new_inode->i_size && - old_inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT) + if (new_inode && S_ISREG(old_inode->i_mode) && new_inode->i_size) filemap_flush(old_inode->i_mapping); /* close the racy window with snapshot create/destroy ioctl */ @@ -8403,12 +8368,6 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry, */ btrfs_pin_log_trans(root); } - /* - * make sure the inode gets flushed if it is replacing - * something. - */ - if (new_inode && new_inode->i_size && S_ISREG(old_inode->i_mode)) - btrfs_add_ordered_operation(trans, root, old_inode); inode_inc_iversion(old_dir); inode_inc_iversion(new_dir); diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 7187b14..963895c 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -571,18 +571,6 @@ void btrfs_remove_ordered_extent(struct inode *inode, trace_btrfs_ordered_extent_remove(inode, entry); - /* - * we have no more ordered extents for this inode and - * no dirty pages. We can safely remove it from the - * list of ordered extents - */ - if (RB_EMPTY_ROOT(&tree->tree) && - !mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY)) { - spin_lock(&root->fs_info->ordered_root_lock); - list_del_init(&BTRFS_I(inode)->ordered_operations); - spin_unlock(&root->fs_info->ordered_root_lock); - } - if (!root->nr_ordered_extents) { spin_lock(&root->fs_info->ordered_root_lock); BUG_ON(list_empty(&root->ordered_root)); @@ -687,81 +675,6 @@ void btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, int nr) } /* - * this is used during transaction commit to write all the inodes - * added to the ordered operation list. These files must be fully on - * disk before the transaction commits. - * - * we have two modes here, one is to just start the IO via filemap_flush - * and the other is to wait for all the io. When we wait, we have an - * extra check to make sure the ordered operation list really is empty - * before we return - */ -int btrfs_run_ordered_operations(struct btrfs_trans_handle *trans, - struct btrfs_root *root, int wait) -{ - struct btrfs_inode *btrfs_inode; - struct inode *inode; - struct btrfs_transaction *cur_trans = trans->transaction; - struct list_head splice; - struct list_head works; - struct btrfs_delalloc_work *work, *next; - int ret = 0; - - INIT_LIST_HEAD(&splice); - INIT_LIST_HEAD(&works); - - mutex_lock(&root->fs_info->ordered_extent_flush_mutex); - spin_lock(&root->fs_info->ordered_root_lock); - list_splice_init(&cur_trans->ordered_operations, &splice); - while (!list_empty(&splice)) { - btrfs_inode = list_entry(splice.next, struct btrfs_inode, - ordered_operations); - inode = &btrfs_inode->vfs_inode; - - list_del_init(&btrfs_inode->ordered_operations); - - /* - * the inode may be getting freed (in sys_unlink path). - */ - inode = igrab(inode); - if (!inode) - continue; - - if (!wait) - list_add_tail(&BTRFS_I(inode)->ordered_operations, - &cur_trans->ordered_operations); - spin_unlock(&root->fs_info->ordered_root_lock); - - work = btrfs_alloc_delalloc_work(inode, wait, 1); - if (!work) { - spin_lock(&root->fs_info->ordered_root_lock); - if (list_empty(&BTRFS_I(inode)->ordered_operations)) - list_add_tail(&btrfs_inode->ordered_operations, - &splice); - list_splice_tail(&splice, - &cur_trans->ordered_operations); - spin_unlock(&root->fs_info->ordered_root_lock); - ret = -ENOMEM; - goto out; - } - list_add_tail(&work->list, &works); - btrfs_queue_work(root->fs_info->flush_workers, - &work->work); - - cond_resched(); - spin_lock(&root->fs_info->ordered_root_lock); - } - spin_unlock(&root->fs_info->ordered_root_lock); -out: - list_for_each_entry_safe(work, next, &works, list) { - list_del_init(&work->list); - btrfs_wait_and_free_delalloc_work(work); - } - mutex_unlock(&root->fs_info->ordered_extent_flush_mutex); - return ret; -} - -/* * Used to start IO or wait for a given ordered extent to finish. * * If wait is one, this effectively waits on page writeback for all the pages @@ -1120,42 +1033,6 @@ out: return index; } - -/* - * add a given inode to the list of inodes that must be fully on - * disk before a transaction commit finishes. - * - * This basically gives us the ext3 style data=ordered mode, and it is mostly - * used to make sure renamed files are fully on disk. - * - * It is a noop if the inode is already fully on disk. - * - * If trans is not null, we'll do a friendly check for a transaction that - * is already flushing things and force the IO down ourselves. - */ -void btrfs_add_ordered_operation(struct btrfs_trans_handle *trans, - struct btrfs_root *root, struct inode *inode) -{ - struct btrfs_transaction *cur_trans = trans->transaction; - u64 last_mod; - - last_mod = max(BTRFS_I(inode)->generation, BTRFS_I(inode)->last_trans); - - /* - * if this file hasn't been changed since the last transaction - * commit, we can safely return without doing anything - */ - if (last_mod <= root->fs_info->last_trans_committed) - return; - - spin_lock(&root->fs_info->ordered_root_lock); - if (list_empty(&BTRFS_I(inode)->ordered_operations)) { - list_add_tail(&BTRFS_I(inode)->ordered_operations, - &cur_trans->ordered_operations); - } - spin_unlock(&root->fs_info->ordered_root_lock); -} - int __init ordered_data_init(void) { btrfs_ordered_extent_cache = kmem_cache_create("btrfs_ordered_extent", diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 2468970..d81a274 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -190,11 +190,6 @@ int btrfs_ordered_update_i_size(struct inode *inode, u64 offset, struct btrfs_ordered_extent *ordered); int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr, u32 *sum, int len); -int btrfs_run_ordered_operations(struct btrfs_trans_handle *trans, - struct btrfs_root *root, int wait); -void btrfs_add_ordered_operation(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct inode *inode); int btrfs_wait_ordered_extents(struct btrfs_root *root, int nr); void btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, int nr); void btrfs_get_logged_extents(struct inode *inode, diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 5f379af..d89c6d3 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -218,7 +218,6 @@ loop: spin_lock_init(&cur_trans->delayed_refs.lock); INIT_LIST_HEAD(&cur_trans->pending_snapshots); - INIT_LIST_HEAD(&cur_trans->ordered_operations); INIT_LIST_HEAD(&cur_trans->pending_chunks); INIT_LIST_HEAD(&cur_trans->switch_commits); list_add_tail(&cur_trans->list, &fs_info->trans_list); @@ -1612,27 +1611,6 @@ static void cleanup_transaction(struct btrfs_trans_handle *trans, kmem_cache_free(btrfs_trans_handle_cachep, trans); } -static int btrfs_flush_all_pending_stuffs(struct btrfs_trans_handle *trans, - struct btrfs_root *root) -{ - int ret; - - ret = btrfs_run_delayed_items(trans, root); - if (ret) - return ret; - - /* - * rename don't use btrfs_join_transaction, so, once we - * set the transaction to blocked above, we aren't going - * to get any new ordered operations. We can safely run - * it here and no for sure that nothing new will be added - * to the list - */ - ret = btrfs_run_ordered_operations(trans, root, 1); - - return ret; -} - static inline int btrfs_start_delalloc_flush(struct btrfs_fs_info *fs_info) { if (btrfs_test_opt(fs_info->tree_root, FLUSHONCOMMIT)) @@ -1653,13 +1631,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_transaction *prev_trans = NULL; int ret; - ret = btrfs_run_ordered_operations(trans, root, 0); - if (ret) { - btrfs_abort_transaction(trans, root, ret); - btrfs_end_transaction(trans, root); - return ret; - } - /* Stop the commit early if ->aborted is set */ if (unlikely(ACCESS_ONCE(cur_trans->aborted))) { ret = cur_trans->aborted; @@ -1740,7 +1711,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, if (ret) goto cleanup_transaction; - ret = btrfs_flush_all_pending_stuffs(trans, root); + ret = btrfs_run_delayed_items(trans, root); if (ret) goto cleanup_transaction; @@ -1748,7 +1719,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, extwriter_counter_read(cur_trans) == 0); /* some pending stuffs might be added after the previous flush. */ - ret = btrfs_flush_all_pending_stuffs(trans, root); + ret = btrfs_run_delayed_items(trans, root); if (ret) goto cleanup_transaction; diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 7dd558e..579be51 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -55,7 +55,6 @@ struct btrfs_transaction { wait_queue_head_t writer_wait; wait_queue_head_t commit_wait; struct list_head pending_snapshots; - struct list_head ordered_operations; struct list_head pending_chunks; struct list_head switch_commits; struct btrfs_delayed_ref_root delayed_refs; -- cgit v1.1