1 files changed, 135 insertions, 100 deletions

diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
index 13b94d2e605b..13f1d2e91540 100644
--- a/fs/xfs/xfs_log_recover.c
+++ b/fs/xfs/xfs_log_recover.c
@@ -361,7 +361,7 @@ xlog_find_verify_cycle(
 	*new_blk = -1;
 
 out:
-	kmem_free(buffer);
+	kvfree(buffer);
 	return error;
 }
 
@@ -477,7 +477,7 @@ xlog_find_verify_log_record(
 		*last_blk = i;
 
 out:
-	kmem_free(buffer);
+	kvfree(buffer);
 	return error;
 }
 
@@ -731,7 +731,7 @@ validate_head:
 			goto out_free_buffer;
 	}
 
-	kmem_free(buffer);
+	kvfree(buffer);
 	if (head_blk == log_bbnum)
 		*return_head_blk = 0;
 	else
@@ -745,7 +745,7 @@ validate_head:
 	return 0;
 
 out_free_buffer:
-	kmem_free(buffer);
+	kvfree(buffer);
 	if (error)
 		xfs_warn(log->l_mp, "failed to find log head");
 	return error;
@@ -999,7 +999,7 @@ xlog_verify_tail(
 		"Tail block (0x%llx) overwrite detected. Updated to 0x%llx",
 			 orig_tail, *tail_blk);
 out:
-	kmem_free(buffer);
+	kvfree(buffer);
 	return error;
 }
 
@@ -1046,7 +1046,7 @@ xlog_verify_head(
 	error = xlog_rseek_logrec_hdr(log, *head_blk, *tail_blk,
 				      XLOG_MAX_ICLOGS, tmp_buffer,
 				      &tmp_rhead_blk, &tmp_rhead, &tmp_wrapped);
-	kmem_free(tmp_buffer);
+	kvfree(tmp_buffer);
 	if (error < 0)
 		return error;
 
@@ -1365,7 +1365,7 @@ xlog_find_tail(
 		error = xlog_clear_stale_blocks(log, tail_lsn);
 
 done:
-	kmem_free(buffer);
+	kvfree(buffer);
 
 	if (error)
 		xfs_warn(log->l_mp, "failed to locate log tail");
@@ -1399,6 +1399,7 @@ xlog_find_zeroed(
 	xfs_daddr_t	new_blk, last_blk, start_blk;
 	xfs_daddr_t     num_scan_bblks;
 	int	        error, log_bbnum = log->l_logBBsize;
+	int		ret = 1;
 
 	*blk_no = 0;
 
@@ -1413,8 +1414,7 @@ xlog_find_zeroed(
 	first_cycle = xlog_get_cycle(offset);
 	if (first_cycle == 0) {		/* completely zeroed log */
 		*blk_no = 0;
-		kmem_free(buffer);
-		return 1;
+		goto out_free_buffer;
 	}
 
 	/* check partially zeroed log */
@@ -1424,8 +1424,8 @@ xlog_find_zeroed(
 
 	last_cycle = xlog_get_cycle(offset);
 	if (last_cycle != 0) {		/* log completely written to */
-		kmem_free(buffer);
-		return 0;
+		ret = 0;
+		goto out_free_buffer;
 	}
 
 	/* we have a partially zeroed log */
@@ -1471,10 +1471,10 @@ xlog_find_zeroed(
 
 	*blk_no = last_blk;
 out_free_buffer:
-	kmem_free(buffer);
+	kvfree(buffer);
 	if (error)
 		return error;
-	return 1;
+	return ret;
 }
 
 /*
@@ -1583,7 +1583,7 @@ xlog_write_log_records(
 	}
 
 out_free_buffer:
-	kmem_free(buffer);
+	kvfree(buffer);
 	return error;
 }
 
@@ -1723,30 +1723,24 @@ xlog_clear_stale_blocks(
  */
 void
 xlog_recover_release_intent(
-	struct xlog		*log,
-	unsigned short		intent_type,
-	uint64_t		intent_id)
+	struct xlog			*log,
+	unsigned short			intent_type,
+	uint64_t			intent_id)
 {
-	struct xfs_ail_cursor	cur;
-	struct xfs_log_item	*lip;
-	struct xfs_ail		*ailp = log->l_ailp;
+	struct xfs_defer_pending	*dfp, *n;
+
+	list_for_each_entry_safe(dfp, n, &log->r_dfops, dfp_list) {
+		struct xfs_log_item	*lip = dfp->dfp_intent;
 
-	spin_lock(&ailp->ail_lock);
-	for (lip = xfs_trans_ail_cursor_first(ailp, &cur, 0); lip != NULL;
-	     lip = xfs_trans_ail_cursor_next(ailp, &cur)) {
 		if (lip->li_type != intent_type)
 			continue;
 		if (!lip->li_ops->iop_match(lip, intent_id))
 			continue;
 
-		spin_unlock(&ailp->ail_lock);
-		lip->li_ops->iop_release(lip);
-		spin_lock(&ailp->ail_lock);
-		break;
-	}
+		ASSERT(xlog_item_is_intent(lip));
 
-	xfs_trans_ail_cursor_done(&cur);
-	spin_unlock(&ailp->ail_lock);
+		xfs_defer_cancel_recovery(log->l_mp, dfp);
+	}
 }
 
 int
@@ -1939,6 +1933,29 @@ xlog_buf_readahead(
 		xfs_buf_readahead(log->l_mp->m_ddev_targp, blkno, len, ops);
 }
 
+/*
+ * Create a deferred work structure for resuming and tracking the progress of a
+ * log intent item that was found during recovery.
+ */
+void
+xlog_recover_intent_item(
+	struct xlog			*log,
+	struct xfs_log_item		*lip,
+	xfs_lsn_t			lsn,
+	const struct xfs_defer_op_type	*ops)
+{
+	ASSERT(xlog_item_is_intent(lip));
+
+	xfs_defer_start_recovery(lip, &log->r_dfops, ops);
+
+	/*
+	 * Insert the intent into the AIL directly and drop one reference so
+	 * that finishing or canceling the work will drop the other.
+	 */
+	xfs_trans_ail_insert(log->l_ailp, lip, lsn);
+	lip->li_ops->iop_unpin(lip, 0);
+}
+
 STATIC int
 xlog_recover_items_pass2(
 	struct xlog                     *log,
@@ -2040,7 +2057,8 @@ xlog_recover_add_item(
 {
 	struct xlog_recover_item *item;
 
-	item = kmem_zalloc(sizeof(struct xlog_recover_item), 0);
+	item = kzalloc(sizeof(struct xlog_recover_item),
+			GFP_KERNEL | __GFP_NOFAIL);
 	INIT_LIST_HEAD(&item->ri_list);
 	list_add_tail(&item->ri_list, head);
 }
@@ -2143,7 +2161,7 @@ xlog_recover_add_to_trans(
 		return 0;
 	}
 
-	ptr = kmem_alloc(len, 0);
+	ptr = xlog_kvmalloc(len);
 	memcpy(ptr, dp, len);
 	in_f = (struct xfs_inode_log_format *)ptr;
 
@@ -2165,14 +2183,13 @@ xlog_recover_add_to_trans(
 		"bad number of regions (%d) in inode log format",
 				  in_f->ilf_size);
 			ASSERT(0);
-			kmem_free(ptr);
+			kvfree(ptr);
 			return -EFSCORRUPTED;
 		}
 
 		item->ri_total = in_f->ilf_size;
-		item->ri_buf =
-			kmem_zalloc(item->ri_total * sizeof(xfs_log_iovec_t),
-				    0);
+		item->ri_buf = kzalloc(item->ri_total * sizeof(xfs_log_iovec_t),
+				GFP_KERNEL | __GFP_NOFAIL);
 	}
 
 	if (item->ri_total <= item->ri_cnt) {
@@ -2180,7 +2197,7 @@ xlog_recover_add_to_trans(
 	"log item region count (%d) overflowed size (%d)",
 				item->ri_cnt, item->ri_total);
 		ASSERT(0);
-		kmem_free(ptr);
+		kvfree(ptr);
 		return -EFSCORRUPTED;
 	}
 
@@ -2210,13 +2227,13 @@ xlog_recover_free_trans(
 		/* Free the regions in the item. */
 		list_del(&item->ri_list);
 		for (i = 0; i < item->ri_cnt; i++)
-			kmem_free(item->ri_buf[i].i_addr);
+			kvfree(item->ri_buf[i].i_addr);
 		/* Free the item itself */
-		kmem_free(item->ri_buf);
-		kmem_free(item);
+		kfree(item->ri_buf);
+		kfree(item);
 	}
 	/* Free the transaction recover structure */
-	kmem_free(trans);
+	kfree(trans);
 }
 
 /*
@@ -2315,7 +2332,7 @@ xlog_recover_ophdr_to_trans(
 	 * This is a new transaction so allocate a new recovery container to
 	 * hold the recovery ops that will follow.
 	 */
-	trans = kmem_zalloc(sizeof(struct xlog_recover), 0);
+	trans = kzalloc(sizeof(struct xlog_recover), GFP_KERNEL | __GFP_NOFAIL);
 	trans->r_log_tid = tid;
 	trans->r_lsn = be64_to_cpu(rhead->h_lsn);
 	INIT_LIST_HEAD(&trans->r_itemq);
@@ -2511,7 +2528,7 @@ xlog_abort_defer_ops(
 
 	list_for_each_entry_safe(dfc, next, capture_list, dfc_list) {
 		list_del_init(&dfc->dfc_list);
-		xfs_defer_ops_capture_free(mp, dfc);
+		xfs_defer_ops_capture_abort(mp, dfc);
 	}
 }
 
@@ -2533,36 +2550,26 @@ xlog_abort_defer_ops(
  */
 STATIC int
 xlog_recover_process_intents(
-	struct xlog		*log)
+	struct xlog			*log)
 {
 	LIST_HEAD(capture_list);
-	struct xfs_ail_cursor	cur;
-	struct xfs_log_item	*lip;
-	struct xfs_ail		*ailp;
-	int			error = 0;
+	struct xfs_defer_pending	*dfp, *n;
+	int				error = 0;
 #if defined(DEBUG) || defined(XFS_WARN)
-	xfs_lsn_t		last_lsn;
-#endif
+	xfs_lsn_t			last_lsn;
 
-	ailp = log->l_ailp;
-	spin_lock(&ailp->ail_lock);
-#if defined(DEBUG) || defined(XFS_WARN)
 	last_lsn = xlog_assign_lsn(log->l_curr_cycle, log->l_curr_block);
 #endif
-	for (lip = xfs_trans_ail_cursor_first(ailp, &cur, 0);
-	     lip != NULL;
-	     lip = xfs_trans_ail_cursor_next(ailp, &cur)) {
-		const struct xfs_item_ops	*ops;
 
-		if (!xlog_item_is_intent(lip))
-			break;
+	list_for_each_entry_safe(dfp, n, &log->r_dfops, dfp_list) {
+		ASSERT(xlog_item_is_intent(dfp->dfp_intent));
 
 		/*
 		 * We should never see a redo item with a LSN higher than
 		 * the last transaction we found in the log at the start
 		 * of recovery.
 		 */
-		ASSERT(XFS_LSN_CMP(last_lsn, lip->li_lsn) >= 0);
+		ASSERT(XFS_LSN_CMP(last_lsn, dfp->dfp_intent->li_lsn) >= 0);
 
 		/*
 		 * NOTE: If your intent processing routine can create more
@@ -2571,21 +2578,14 @@ xlog_recover_process_intents(
 		 * replayed in the wrong order!
 		 *
 		 * The recovery function can free the log item, so we must not
-		 * access lip after it returns.
+		 * access dfp->dfp_intent after it returns.  It must dispose of
+		 * @dfp if it returns 0.
 		 */
-		spin_unlock(&ailp->ail_lock);
-		ops = lip->li_ops;
-		error = ops->iop_recover(lip, &capture_list);
-		spin_lock(&ailp->ail_lock);
-		if (error) {
-			trace_xlog_intent_recovery_failed(log->l_mp, error,
-					ops->iop_recover);
+		error = xfs_defer_finish_recovery(log->l_mp, dfp,
+				&capture_list);
+		if (error)
 			break;
-		}
 	}
-
-	xfs_trans_ail_cursor_done(&cur);
-	spin_unlock(&ailp->ail_lock);
 	if (error)
 		goto err;
 
@@ -2606,27 +2606,34 @@ err:
  */
 STATIC void
 xlog_recover_cancel_intents(
-	struct xlog		*log)
+	struct xlog			*log)
 {
-	struct xfs_log_item	*lip;
-	struct xfs_ail_cursor	cur;
-	struct xfs_ail		*ailp;
-
-	ailp = log->l_ailp;
-	spin_lock(&ailp->ail_lock);
-	lip = xfs_trans_ail_cursor_first(ailp, &cur, 0);
-	while (lip != NULL) {
-		if (!xlog_item_is_intent(lip))
-			break;
+	struct xfs_defer_pending	*dfp, *n;
 
-		spin_unlock(&ailp->ail_lock);
-		lip->li_ops->iop_release(lip);
-		spin_lock(&ailp->ail_lock);
-		lip = xfs_trans_ail_cursor_next(ailp, &cur);
+	list_for_each_entry_safe(dfp, n, &log->r_dfops, dfp_list) {
+		ASSERT(xlog_item_is_intent(dfp->dfp_intent));
+
+		xfs_defer_cancel_recovery(log->l_mp, dfp);
 	}
+}
+
+/*
+ * Transfer ownership of the recovered pending work to the recovery transaction
+ * and try to finish the work.  If there is more work to be done, the dfp will
+ * remain attached to the transaction.  If not, the dfp is freed.
+ */
+int
+xlog_recover_finish_intent(
+	struct xfs_trans		*tp,
+	struct xfs_defer_pending	*dfp)
+{
+	int				error;
 
-	xfs_trans_ail_cursor_done(&cur);
-	spin_unlock(&ailp->ail_lock);
+	list_move(&dfp->dfp_list, &tp->t_dfops);
+	error = xfs_defer_finish_one(tp, dfp);
+	if (error == -EAGAIN)
+		return 0;
+	return error;
 }
 
 /*
@@ -3017,7 +3024,7 @@ xlog_do_recovery_pass(
 
 		hblks = xlog_logrec_hblks(log, rhead);
 		if (hblks != 1) {
-			kmem_free(hbp);
+			kvfree(hbp);
 			hbp = xlog_alloc_buffer(log, hblks);
 		}
 	} else {
@@ -3031,7 +3038,7 @@ xlog_do_recovery_pass(
 		return -ENOMEM;
 	dbp = xlog_alloc_buffer(log, BTOBB(h_size));
 	if (!dbp) {
-		kmem_free(hbp);
+		kvfree(hbp);
 		return -ENOMEM;
 	}
 
@@ -3192,16 +3199,33 @@ xlog_do_recovery_pass(
 	}
 
  bread_err2:
-	kmem_free(dbp);
+	kvfree(dbp);
  bread_err1:
-	kmem_free(hbp);
+	kvfree(hbp);
 
 	/*
-	 * Submit buffers that have been added from the last record processed,
-	 * regardless of error status.
+	 * Submit buffers that have been dirtied by the last record recovered.
 	 */
-	if (!list_empty(&buffer_list))
+	if (!list_empty(&buffer_list)) {
+		if (error) {
+			/*
+			 * If there has been an item recovery error then we
+			 * cannot allow partial checkpoint writeback to
+			 * occur.  We might have multiple checkpoints with the
+			 * same start LSN in this buffer list, and partial
+			 * writeback of a checkpoint in this situation can
+			 * prevent future recovery of all the changes in the
+			 * checkpoints at this start LSN.
+			 *
+			 * Note: Shutting down the filesystem will result in the
+			 * delwri submission marking all the buffers stale,
+			 * completing them and cleaning up _XBF_LOGRECOVERY
+			 * state without doing any IO.
+			 */
+			xlog_force_shutdown(log, SHUTDOWN_LOG_IO_ERROR);
+		}
 		error2 = xfs_buf_delwri_submit(&buffer_list);
+	}
 
 	if (error && first_bad)
 		*first_bad = rhead_blk;
@@ -3436,12 +3460,19 @@ xlog_recover(
  * part of recovery so that the root and real-time bitmap inodes can be read in
  * from disk in between the two stages.  This is necessary so that we can free
  * space in the real-time portion of the file system.
+ *
+ * We run this whole process under GFP_NOFS allocation context. We do a
+ * combination of non-transactional and transactional work, yet we really don't
+ * want to recurse into the filesystem from direct reclaim during any of this
+ * processing. This allows all the recovery code run here not to care about the
+ * memory allocation context it is running in.
  */
 int
 xlog_recover_finish(
 	struct xlog	*log)
 {
-	int	error;
+	unsigned int	nofs_flags = memalloc_nofs_save();
+	int		error;
 
 	error = xlog_recover_process_intents(log);
 	if (error) {
@@ -3455,7 +3486,7 @@ xlog_recover_finish(
 		xlog_recover_cancel_intents(log);
 		xfs_alert(log->l_mp, "Failed to recover intents");
 		xlog_force_shutdown(log, SHUTDOWN_LOG_IO_ERROR);
-		return error;
+		goto out_error;
 	}
 
 	/*
@@ -3476,7 +3507,7 @@ xlog_recover_finish(
 		if (error < 0) {
 			xfs_alert(log->l_mp,
 	"Failed to clear log incompat features on recovery");
-			return error;
+			goto out_error;
 		}
 	}
 
@@ -3501,9 +3532,13 @@ xlog_recover_finish(
 		 * and AIL.
 		 */
 		xlog_force_shutdown(log, SHUTDOWN_LOG_IO_ERROR);
+		error = 0;
+		goto out_error;
 	}
 
-	return 0;
+out_error:
+	memalloc_nofs_restore(nofs_flags);
+	return error;
 }
 
 void