xfs: proper replay of deferred ops queued during log recovery

When we replay unfinished intent items that have been recovered from the
log, it's possible that the replay will cause the creation of more
deferred work items.  As outlined in commit 509955823cc9c ("xfs: log
recovery should replay deferred ops in order"), later work items have an
implicit ordering dependency on earlier work items.  Therefore, recovery
must replay the items (both recovered and created) in the same order
that they would have been during normal operation.

For log recovery, we enforce this ordering by using an empty transaction
to collect deferred ops that get created in the process of recovering a
log intent item to prevent them from being committed before the rest of
the recovered intent items.  After we finish committing all the
recovered log items, we allocate a transaction with an enormous block
reservation, splice our huge list of created deferred ops into that
transaction, and commit it, thereby finishing all those ops.

This is /really/ hokey -- it's the one place in XFS where we allow
nested transactions; the splicing of the defer ops list is is inelegant
and has to be done twice per recovery function; and the broken way we
handle inode pointers and block reservations cause subtle use-after-free
and allocator problems that will be fixed by this patch and the two
patches after it.

Therefore, replace the hokey empty transaction with a structure designed
to capture each chain of deferred ops that are created as part of
recovering a single unfinished log intent.  Finally, refactor the loop
that replays those chains to do so using one transaction per chain.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

1 files changed, 71 insertions, 50 deletions

diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
index 84f876c6d498..7804906d145b 100644
--- a/fs/xfs/xfs_log_recover.c
+++ b/fs/xfs/xfs_log_recover.c
@@ -2434,37 +2434,65 @@ xlog_recover_process_data(
 /* Take all the collected deferred ops and finish them in order. */
 static int
 xlog_finish_defer_ops(
-	struct xfs_trans	*parent_tp)
+	struct xfs_mount	*mp,
+	struct list_head	*capture_list)
 {
-	struct xfs_mount	*mp = parent_tp->t_mountp;
+	struct xfs_defer_capture *dfc, *next;
 	struct xfs_trans	*tp;
 	int64_t			freeblks;
-	uint			resblks;
-	int			error;
+	uint64_t		resblks;
+	int			error = 0;
 
-	/*
-	 * We're finishing the defer_ops that accumulated as a result of
-	 * recovering unfinished intent items during log recovery.  We
-	 * reserve an itruncate transaction because it is the largest
-	 * permanent transaction type.  Since we're the only user of the fs
-	 * right now, take 93% (15/16) of the available free blocks.  Use
-	 * weird math to avoid a 64-bit division.
-	 */
-	freeblks = percpu_counter_sum(&mp->m_fdblocks);
-	if (freeblks <= 0)
-		return -ENOSPC;
-	resblks = min_t(int64_t, UINT_MAX, freeblks);
-	resblks = (resblks * 15) >> 4;
-	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, resblks,
-			0, XFS_TRANS_RESERVE, &tp);
-	if (error)
-		return error;
-	/* transfer all collected dfops to this transaction */
-	xfs_defer_move(tp, parent_tp);
+	list_for_each_entry_safe(dfc, next, capture_list, dfc_list) {
+		/*
+		 * We're finishing the defer_ops that accumulated as a result
+		 * of recovering unfinished intent items during log recovery.
+		 * We reserve an itruncate transaction because it is the
+		 * largest permanent transaction type.  Since we're the only
+		 * user of the fs right now, take 93% (15/16) of the available
+		 * free blocks.  Use weird math to avoid a 64-bit division.
+		 */
+		freeblks = percpu_counter_sum(&mp->m_fdblocks);
+		if (freeblks <= 0)
+			return -ENOSPC;
+
+		resblks = min_t(uint64_t, UINT_MAX, freeblks);
+		resblks = (resblks * 15) >> 4;
+		error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, resblks,
+				0, XFS_TRANS_RESERVE, &tp);
+		if (error)
+			return error;
+
+		/*
+		 * Transfer to this new transaction all the dfops we captured
+		 * from recovering a single intent item.
+		 */
+		list_del_init(&dfc->dfc_list);
+		xfs_defer_ops_continue(dfc, tp);
+
+		error = xfs_trans_commit(tp);
+		if (error)
+			return error;
+	}
 
-	return xfs_trans_commit(tp);
+	ASSERT(list_empty(capture_list));
+	return 0;
 }
 
+/* Release all the captured defer ops and capture structures in this list. */
+static void
+xlog_abort_defer_ops(
+	struct xfs_mount		*mp,
+	struct list_head		*capture_list)
+{
+	struct xfs_defer_capture	*dfc;
+	struct xfs_defer_capture	*next;
+
+	list_for_each_entry_safe(dfc, next, capture_list, dfc_list) {
+		list_del_init(&dfc->dfc_list);
+		xfs_defer_ops_release(mp, dfc);
+	}
+}
 /*
  * When this is called, all of the log intent items which did not have
  * corresponding log done items should be in the AIL.  What we do now
@@ -2485,35 +2513,23 @@ STATIC int
 xlog_recover_process_intents(
 	struct xlog		*log)
 {
-	struct xfs_trans	*parent_tp;
+	LIST_HEAD(capture_list);
 	struct xfs_ail_cursor	cur;
 	struct xfs_log_item	*lip;
 	struct xfs_ail		*ailp;
-	int			error;
+	int			error = 0;
 #if defined(DEBUG) || defined(XFS_WARN)
 	xfs_lsn_t		last_lsn;
 #endif
 
-	/*
-	 * The intent recovery handlers commit transactions to complete recovery
-	 * for individual intents, but any new deferred operations that are
-	 * queued during that process are held off until the very end. The
-	 * purpose of this transaction is to serve as a container for deferred
-	 * operations. Each intent recovery handler must transfer dfops here
-	 * before its local transaction commits, and we'll finish the entire
-	 * list below.
-	 */
-	error = xfs_trans_alloc_empty(log->l_mp, &parent_tp);
-	if (error)
-		return error;
-
 	ailp = log->l_ailp;
 	spin_lock(&ailp->ail_lock);
-	lip = xfs_trans_ail_cursor_first(ailp, &cur, 0);
 #if defined(DEBUG) || defined(XFS_WARN)
 	last_lsn = xlog_assign_lsn(log->l_curr_cycle, log->l_curr_block);
 #endif
-	while (lip != NULL) {
+	for (lip = xfs_trans_ail_cursor_first(ailp, &cur, 0);
+	     lip != NULL;
+	     lip = xfs_trans_ail_cursor_next(ailp, &cur)) {
 		/*
 		 * We're done when we see something other than an intent.
 		 * There should be no intents left in the AIL now.
@@ -2535,24 +2551,29 @@ xlog_recover_process_intents(
 
 		/*
 		 * NOTE: If your intent processing routine can create more
-		 * deferred ops, you /must/ attach them to the transaction in
-		 * this routine or else those subsequent intents will get
+		 * deferred ops, you /must/ attach them to the capture list in
+		 * the recover routine or else those subsequent intents will be
 		 * replayed in the wrong order!
 		 */
 		spin_unlock(&ailp->ail_lock);
-		error = lip->li_ops->iop_recover(lip, parent_tp);
+		error = lip->li_ops->iop_recover(lip, &capture_list);
 		spin_lock(&ailp->ail_lock);
 		if (error)
-			goto out;
-		lip = xfs_trans_ail_cursor_next(ailp, &cur);
+			break;
 	}
-out:
+
 	xfs_trans_ail_cursor_done(&cur);
 	spin_unlock(&ailp->ail_lock);
-	if (!error)
-		error = xlog_finish_defer_ops(parent_tp);
-	xfs_trans_cancel(parent_tp);
+	if (error)
+		goto err;
 
+	error = xlog_finish_defer_ops(log->l_mp, &capture_list);
+	if (error)
+		goto err;
+
+	return 0;
+err:
+	xlog_abort_defer_ops(log->l_mp, &capture_list);
 	return error;
 }