13 files changed, 230 insertions, 28 deletions

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 947f04789389..ebb2d109e8bb 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -602,6 +602,9 @@ enum {
 	/* Indicate that we want the transaction kthread to commit right now. */
 	BTRFS_FS_COMMIT_TRANS,
 
+	/* Indicate we have half completed snapshot deletions pending. */
+	BTRFS_FS_UNFINISHED_DROPS,
+
 #if BITS_PER_LONG == 32
 	/* Indicate if we have error/warn message printed on 32bit systems */
 	BTRFS_FS_32BIT_ERROR,
@@ -1106,8 +1109,15 @@ enum {
 	BTRFS_ROOT_QGROUP_FLUSHING,
 	/* We started the orphan cleanup for this root. */
 	BTRFS_ROOT_ORPHAN_CLEANUP,
+	/* This root has a drop operation that was started previously. */
+	BTRFS_ROOT_UNFINISHED_DROP,
 };
 
+static inline void btrfs_wake_unfinished_drop(struct btrfs_fs_info *fs_info)
+{
+	clear_and_wake_up_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags);
+}
+
 /*
  * Record swapped tree blocks of a subvolume tree for delayed subtree trace
  * code. For detail check comment in fs/btrfs/qgroup.c.
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 87a5addbedf6..48590a380762 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -3813,6 +3813,10 @@ int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_device
 
 	set_bit(BTRFS_FS_OPEN, &fs_info->flags);
 
+	/* Kick the cleaner thread so it'll start deleting snapshots. */
+	if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags))
+		wake_up_process(fs_info->cleaner_kthread);
+
 clear_oneshot:
 	btrfs_clear_oneshot_options(fs_info);
 	return 0;
@@ -4538,6 +4542,12 @@ void __cold close_ctree(struct btrfs_fs_info *fs_info)
 	 */
 	kthread_park(fs_info->cleaner_kthread);
 
+	/*
+	 * If we had UNFINISHED_DROPS we could still be processing them, so
+	 * clear that bit and wake up relocation so it can stop.
+	 */
+	btrfs_wake_unfinished_drop(fs_info);
+
 	/* wait for the qgroup rescan worker to stop */
 	btrfs_qgroup_wait_for_completion(fs_info, false);
 
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index d89273c4b6b8..96427b1ecac3 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -5622,6 +5622,7 @@ int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref, int for_reloc)
 	int ret;
 	int level;
 	bool root_dropped = false;
+	bool unfinished_drop = false;
 
 	btrfs_debug(fs_info, "Drop subvolume %llu", root->root_key.objectid);
 
@@ -5664,6 +5665,8 @@ int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref, int for_reloc)
 	 * already dropped.
 	 */
 	set_bit(BTRFS_ROOT_DELETING, &root->state);
+	unfinished_drop = test_bit(BTRFS_ROOT_UNFINISHED_DROP, &root->state);
+
 	if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
 		level = btrfs_header_level(root->node);
 		path->nodes[level] = btrfs_lock_root_node(root);
@@ -5839,6 +5842,13 @@ out_free:
 	btrfs_free_path(path);
 out:
 	/*
+	 * We were an unfinished drop root, check to see if there are any
+	 * pending, and if not clear and wake up any waiters.
+	 */
+	if (!err && unfinished_drop)
+		btrfs_maybe_wake_unfinished_drop(fs_info);
+
+	/*
 	 * So if we need to stop dropping the snapshot for whatever reason we
 	 * need to make sure to add it back to the dead root list so that we
 	 * keep trying to do the work later.  This also cleans up roots if we
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 409bad3928db..4c91060d103a 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -6841,14 +6841,24 @@ static void assert_eb_page_uptodate(const struct extent_buffer *eb,
 {
 	struct btrfs_fs_info *fs_info = eb->fs_info;
 
+	/*
+	 * If we are using the commit root we could potentially clear a page
+	 * Uptodate while we're using the extent buffer that we've previously
+	 * looked up.  We don't want to complain in this case, as the page was
+	 * valid before, we just didn't write it out.  Instead we want to catch
+	 * the case where we didn't actually read the block properly, which
+	 * would have !PageUptodate && !PageError, as we clear PageError before
+	 * reading.
+	 */
 	if (fs_info->sectorsize < PAGE_SIZE) {
-		bool uptodate;
+		bool uptodate, error;
 
 		uptodate = btrfs_subpage_test_uptodate(fs_info, page,
 						       eb->start, eb->len);
-		WARN_ON(!uptodate);
+		error = btrfs_subpage_test_error(fs_info, page, eb->start, eb->len);
+		WARN_ON(!uptodate && !error);
 	} else {
-		WARN_ON(!PageUptodate(page));
+		WARN_ON(!PageUptodate(page) && !PageError(page));
 	}
 }
 
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 76e530f76e3c..5bbea5ec31fc 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -7600,6 +7600,34 @@ static int btrfs_dio_iomap_begin(struct inode *inode, loff_t start,
 	}
 
 	len = min(len, em->len - (start - em->start));
+
+	/*
+	 * If we have a NOWAIT request and the range contains multiple extents
+	 * (or a mix of extents and holes), then we return -EAGAIN to make the
+	 * caller fallback to a context where it can do a blocking (without
+	 * NOWAIT) request. This way we avoid doing partial IO and returning
+	 * success to the caller, which is not optimal for writes and for reads
+	 * it can result in unexpected behaviour for an application.
+	 *
+	 * When doing a read, because we use IOMAP_DIO_PARTIAL when calling
+	 * iomap_dio_rw(), we can end up returning less data then what the caller
+	 * asked for, resulting in an unexpected, and incorrect, short read.
+	 * That is, the caller asked to read N bytes and we return less than that,
+	 * which is wrong unless we are crossing EOF. This happens if we get a
+	 * page fault error when trying to fault in pages for the buffer that is
+	 * associated to the struct iov_iter passed to iomap_dio_rw(), and we
+	 * have previously submitted bios for other extents in the range, in
+	 * which case iomap_dio_rw() may return us EIOCBQUEUED if not all of
+	 * those bios have completed by the time we get the page fault error,
+	 * which we return back to our caller - we should only return EIOCBQUEUED
+	 * after we have submitted bios for all the extents in the range.
+	 */
+	if ((flags & IOMAP_NOWAIT) && len < length) {
+		free_extent_map(em);
+		ret = -EAGAIN;
+		goto unlock_err;
+	}
+
 	if (write) {
 		ret = btrfs_get_blocks_direct_write(&em, inode, dio_data,
 						    start, len);
diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c
index f12dc687350c..30d42ea655ce 100644
--- a/fs/btrfs/qgroup.c
+++ b/fs/btrfs/qgroup.c
@@ -1197,13 +1197,20 @@ int btrfs_quota_disable(struct btrfs_fs_info *fs_info)
 		goto out;
 
 	/*
+	 * Unlock the qgroup_ioctl_lock mutex before waiting for the rescan worker to
+	 * complete. Otherwise we can deadlock because btrfs_remove_qgroup() needs
+	 * to lock that mutex while holding a transaction handle and the rescan
+	 * worker needs to commit a transaction.
+	 */
+	mutex_unlock(&fs_info->qgroup_ioctl_lock);
+
+	/*
 	 * Request qgroup rescan worker to complete and wait for it. This wait
 	 * must be done before transaction start for quota disable since it may
 	 * deadlock with transaction by the qgroup rescan worker.
 	 */
 	clear_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
 	btrfs_qgroup_wait_for_completion(fs_info, false);
-	mutex_unlock(&fs_info->qgroup_ioctl_lock);
 
 	/*
 	 * 1 For the root item
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index f5465197996d..9d8054839782 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -3960,6 +3960,19 @@ int btrfs_relocate_block_group(struct btrfs_fs_info *fs_info, u64 group_start)
 	int rw = 0;
 	int err = 0;
 
+	/*
+	 * This only gets set if we had a half-deleted snapshot on mount.  We
+	 * cannot allow relocation to start while we're still trying to clean up
+	 * these pending deletions.
+	 */
+	ret = wait_on_bit(&fs_info->flags, BTRFS_FS_UNFINISHED_DROPS, TASK_INTERRUPTIBLE);
+	if (ret)
+		return ret;
+
+	/* We may have been woken up by close_ctree, so bail if we're closing. */
+	if (btrfs_fs_closing(fs_info))
+		return -EINTR;
+
 	bg = btrfs_lookup_block_group(fs_info, group_start);
 	if (!bg)
 		return -ENOENT;
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index 3d68d2dcd83e..ca7426ef61c8 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -278,6 +278,21 @@ int btrfs_find_orphan_roots(struct btrfs_fs_info *fs_info)
 
 		WARN_ON(!test_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state));
 		if (btrfs_root_refs(&root->root_item) == 0) {
+			struct btrfs_key drop_key;
+
+			btrfs_disk_key_to_cpu(&drop_key, &root->root_item.drop_progress);
+			/*
+			 * If we have a non-zero drop_progress then we know we
+			 * made it partly through deleting this snapshot, and
+			 * thus we need to make sure we block any balance from
+			 * happening until this snapshot is completely dropped.
+			 */
+			if (drop_key.objectid != 0 || drop_key.type != 0 ||
+			    drop_key.offset != 0) {
+				set_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags);
+				set_bit(BTRFS_ROOT_UNFINISHED_DROP, &root->state);
+			}
+
 			set_bit(BTRFS_ROOT_DEAD_TREE, &root->state);
 			btrfs_add_dead_root(root);
 		}
diff --git a/fs/btrfs/subpage.c b/fs/btrfs/subpage.c
index 29bd8c7a7706..ef7ae20d2b77 100644
--- a/fs/btrfs/subpage.c
+++ b/fs/btrfs/subpage.c
@@ -736,7 +736,7 @@ void btrfs_page_unlock_writer(struct btrfs_fs_info *fs_info, struct page *page,
 	 * Since we own the page lock, no one else could touch subpage::writers
 	 * and we are safe to do several atomic operations without spinlock.
 	 */
-	if (atomic_read(&subpage->writers))
+	if (atomic_read(&subpage->writers) == 0)
 		/* No writers, locked by plain lock_page() */
 		return unlock_page(page);
 
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index c3cfdfd8de9b..1f1c25db6f6b 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -854,7 +854,37 @@ btrfs_attach_transaction_barrier(struct btrfs_root *root)
 static noinline void wait_for_commit(struct btrfs_transaction *commit,
 				     const enum btrfs_trans_state min_state)
 {
-	wait_event(commit->commit_wait, commit->state >= min_state);
+	struct btrfs_fs_info *fs_info = commit->fs_info;
+	u64 transid = commit->transid;
+	bool put = false;
+
+	while (1) {
+		wait_event(commit->commit_wait, commit->state >= min_state);
+		if (put)
+			btrfs_put_transaction(commit);
+
+		if (min_state < TRANS_STATE_COMPLETED)
+			break;
+
+		/*
+		 * A transaction isn't really completed until all of the
+		 * previous transactions are completed, but with fsync we can
+		 * end up with SUPER_COMMITTED transactions before a COMPLETED
+		 * transaction. Wait for those.
+		 */
+
+		spin_lock(&fs_info->trans_lock);
+		commit = list_first_entry_or_null(&fs_info->trans_list,
+						  struct btrfs_transaction,
+						  list);
+		if (!commit || commit->transid > transid) {
+			spin_unlock(&fs_info->trans_lock);
+			break;
+		}
+		refcount_inc(&commit->use_count);
+		put = true;
+		spin_unlock(&fs_info->trans_lock);
+	}
 }
 
 int btrfs_wait_for_commit(struct btrfs_fs_info *fs_info, u64 transid)
@@ -1320,6 +1350,32 @@ again:
 }
 
 /*
+ * If we had a pending drop we need to see if there are any others left in our
+ * dead roots list, and if not clear our bit and wake any waiters.
+ */
+void btrfs_maybe_wake_unfinished_drop(struct btrfs_fs_info *fs_info)
+{
+	/*
+	 * We put the drop in progress roots at the front of the list, so if the
+	 * first entry doesn't have UNFINISHED_DROP set we can wake everybody
+	 * up.
+	 */
+	spin_lock(&fs_info->trans_lock);
+	if (!list_empty(&fs_info->dead_roots)) {
+		struct btrfs_root *root = list_first_entry(&fs_info->dead_roots,
+							   struct btrfs_root,
+							   root_list);
+		if (test_bit(BTRFS_ROOT_UNFINISHED_DROP, &root->state)) {
+			spin_unlock(&fs_info->trans_lock);
+			return;
+		}
+	}
+	spin_unlock(&fs_info->trans_lock);
+
+	btrfs_wake_unfinished_drop(fs_info);
+}
+
+/*
  * dead roots are old snapshots that need to be deleted.  This allocates
  * a dirty root struct and adds it into the list of dead roots that need to
  * be deleted
@@ -1331,7 +1387,12 @@ void btrfs_add_dead_root(struct btrfs_root *root)
 	spin_lock(&fs_info->trans_lock);
 	if (list_empty(&root->root_list)) {
 		btrfs_grab_root(root);
-		list_add_tail(&root->root_list, &fs_info->dead_roots);
+
+		/* We want to process the partially complete drops first. */
+		if (test_bit(BTRFS_ROOT_UNFINISHED_DROP, &root->state))
+			list_add(&root->root_list, &fs_info->dead_roots);
+		else
+			list_add_tail(&root->root_list, &fs_info->dead_roots);
 	}
 	spin_unlock(&fs_info->trans_lock);
 }
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index 9402d8d94484..ba8a9826eb37 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -216,6 +216,7 @@ int btrfs_wait_for_commit(struct btrfs_fs_info *fs_info, u64 transid);
 
 void btrfs_add_dead_root(struct btrfs_root *root);
 int btrfs_defrag_root(struct btrfs_root *root);
+void btrfs_maybe_wake_unfinished_drop(struct btrfs_fs_info *fs_info);
 int btrfs_clean_one_deleted_snapshot(struct btrfs_root *root);
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans);
 void btrfs_commit_transaction_async(struct btrfs_trans_handle *trans);
diff --git a/fs/btrfs/tree-checker.c b/fs/btrfs/tree-checker.c
index 9fd145f1c4bc..aae5697dde32 100644
--- a/fs/btrfs/tree-checker.c
+++ b/fs/btrfs/tree-checker.c
@@ -1682,6 +1682,7 @@ static int check_leaf(struct extent_buffer *leaf, bool check_item_data)
 	 */
 	for (slot = 0; slot < nritems; slot++) {
 		u32 item_end_expected;
+		u64 item_data_end;
 		int ret;
 
 		btrfs_item_key_to_cpu(leaf, &key, slot);
@@ -1696,6 +1697,8 @@ static int check_leaf(struct extent_buffer *leaf, bool check_item_data)
 			return -EUCLEAN;
 		}
 
+		item_data_end = (u64)btrfs_item_offset(leaf, slot) +
+				btrfs_item_size(leaf, slot);
 		/*
 		 * Make sure the offset and ends are right, remember that the
 		 * item data starts at the end of the leaf and grows towards the
@@ -1706,11 +1709,10 @@ static int check_leaf(struct extent_buffer *leaf, bool check_item_data)
 		else
 			item_end_expected = btrfs_item_offset(leaf,
 								 slot - 1);
-		if (unlikely(btrfs_item_data_end(leaf, slot) != item_end_expected)) {
+		if (unlikely(item_data_end != item_end_expected)) {
 			generic_err(leaf, slot,
-				"unexpected item end, have %u expect %u",
-				btrfs_item_data_end(leaf, slot),
-				item_end_expected);
+				"unexpected item end, have %llu expect %u",
+				item_data_end, item_end_expected);
 			return -EUCLEAN;
 		}
 
@@ -1719,12 +1721,10 @@ static int check_leaf(struct extent_buffer *leaf, bool check_item_data)
 		 * just in case all the items are consistent to each other, but
 		 * all point outside of the leaf.
 		 */
-		if (unlikely(btrfs_item_data_end(leaf, slot) >
-			     BTRFS_LEAF_DATA_SIZE(fs_info))) {
+		if (unlikely(item_data_end > BTRFS_LEAF_DATA_SIZE(fs_info))) {
 			generic_err(leaf, slot,
-			"slot end outside of leaf, have %u expect range [0, %u]",
-				btrfs_item_data_end(leaf, slot),
-				BTRFS_LEAF_DATA_SIZE(fs_info));
+			"slot end outside of leaf, have %llu expect range [0, %u]",
+				item_data_end, BTRFS_LEAF_DATA_SIZE(fs_info));
 			return -EUCLEAN;
 		}
 
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 3ee014c06b82..6bc8834ac8f7 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -1362,6 +1362,15 @@ again:
 						 inode, name, namelen);
 			kfree(name);
 			iput(dir);
+			/*
+			 * Whenever we need to check if a name exists or not, we
+			 * check the subvolume tree. So after an unlink we must
+			 * run delayed items, so that future checks for a name
+			 * during log replay see that the name does not exists
+			 * anymore.
+			 */
+			if (!ret)
+				ret = btrfs_run_delayed_items(trans);
 			if (ret)
 				goto out;
 			goto again;
@@ -1614,6 +1623,15 @@ static noinline int add_inode_ref(struct btrfs_trans_handle *trans,
 				 */
 				if (!ret && inode->i_nlink == 0)
 					inc_nlink(inode);
+				/*
+				 * Whenever we need to check if a name exists or
+				 * not, we check the subvolume tree. So after an
+				 * unlink we must run delayed items, so that future
+				 * checks for a name during log replay see that the
+				 * name does not exists anymore.
+				 */
+				if (!ret)
+					ret = btrfs_run_delayed_items(trans);
 			}
 			if (ret < 0)
 				goto out;
@@ -4635,7 +4653,7 @@ static int log_one_extent(struct btrfs_trans_handle *trans,
 
 /*
  * Log all prealloc extents beyond the inode's i_size to make sure we do not
- * lose them after doing a fast fsync and replaying the log. We scan the
+ * lose them after doing a full/fast fsync and replaying the log. We scan the
  * subvolume's root instead of iterating the inode's extent map tree because
  * otherwise we can log incorrect extent items based on extent map conversion.
  * That can happen due to the fact that extent maps are merged when they
@@ -5414,6 +5432,7 @@ static int copy_inode_items_to_log(struct btrfs_trans_handle *trans,
 				   struct btrfs_log_ctx *ctx,
 				   bool *need_log_inode_item)
 {
+	const u64 i_size = i_size_read(&inode->vfs_inode);
 	struct btrfs_root *root = inode->root;
 	int ins_start_slot = 0;
 	int ins_nr = 0;
@@ -5434,13 +5453,21 @@ again:
 		if (min_key->type > max_key->type)
 			break;
 
-		if (min_key->type == BTRFS_INODE_ITEM_KEY)
+		if (min_key->type == BTRFS_INODE_ITEM_KEY) {
 			*need_log_inode_item = false;
-
-		if ((min_key->type == BTRFS_INODE_REF_KEY ||
-		     min_key->type == BTRFS_INODE_EXTREF_KEY) &&
-		    inode->generation == trans->transid &&
-		    !recursive_logging) {
+		} else if (min_key->type == BTRFS_EXTENT_DATA_KEY &&
+			   min_key->offset >= i_size) {
+			/*
+			 * Extents at and beyond eof are logged with
+			 * btrfs_log_prealloc_extents().
+			 * Only regular files have BTRFS_EXTENT_DATA_KEY keys,
+			 * and no keys greater than that, so bail out.
+			 */
+			break;
+		} else if ((min_key->type == BTRFS_INODE_REF_KEY ||
+			    min_key->type == BTRFS_INODE_EXTREF_KEY) &&
+			   inode->generation == trans->transid &&
+			   !recursive_logging) {
 			u64 other_ino = 0;
 			u64 other_parent = 0;
 
@@ -5471,10 +5498,8 @@ again:
 				btrfs_release_path(path);
 				goto next_key;
 			}
-		}
-
-		/* Skip xattrs, we log them later with btrfs_log_all_xattrs() */
-		if (min_key->type == BTRFS_XATTR_ITEM_KEY) {
+		} else if (min_key->type == BTRFS_XATTR_ITEM_KEY) {
+			/* Skip xattrs, logged later with btrfs_log_all_xattrs() */
 			if (ins_nr == 0)
 				goto next_slot;
 			ret = copy_items(trans, inode, dst_path, path,
@@ -5527,9 +5552,21 @@ next_key:
 			break;
 		}
 	}
-	if (ins_nr)
+	if (ins_nr) {
 		ret = copy_items(trans, inode, dst_path, path, ins_start_slot,
 				 ins_nr, inode_only, logged_isize);
+		if (ret)
+			return ret;
+	}
+
+	if (inode_only == LOG_INODE_ALL && S_ISREG(inode->vfs_inode.i_mode)) {
+		/*
+		 * Release the path because otherwise we might attempt to double
+		 * lock the same leaf with btrfs_log_prealloc_extents() below.
+		 */
+		btrfs_release_path(path);
+		ret = btrfs_log_prealloc_extents(trans, inode, dst_path);
+	}
 
 	return ret;
 }