[PATCH] xfs: reserve blocks for refcount / rmap log item recovery

[PATCH] xfs: reserve blocks for refcount / rmap log item recovery

[Darrick J. Wong]

From: Darrick J. Wong <darrick.wong@oracle.com>

During log recovery, the per-AG reservations aren't yet set up, so log
recovery has to reserve enough blocks to handle all possible btree
splits.

Reported-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/xfs_refcount_item.c |    9 ++++++---
 fs/xfs/xfs_rmap_item.c     |    4 +++-
 2 files changed, 9 insertions(+), 4 deletions(-)

diff --git a/fs/xfs/xfs_refcount_item.c b/fs/xfs/xfs_refcount_item.c
index 3a55d6f..7a39f40 100644
--- a/fs/xfs/xfs_refcount_item.c
+++ b/fs/xfs/xfs_refcount_item.c
@@ -23,6 +23,7 @@
 #include "xfs_log_format.h"
 #include "xfs_trans_resv.h"
 #include "xfs_bit.h"
+#include "xfs_shared.h"
 #include "xfs_mount.h"
 #include "xfs_defer.h"
 #include "xfs_trans.h"
@@ -456,10 +457,12 @@ xfs_cui_recover(
 	 * transaction.  Normally, any work that needs to be deferred
 	 * gets attached to the same defer_ops that scheduled the
 	 * refcount update.  However, we're in log recovery here, so we
-	 * we create our own defer_ops and use that to finish up any
-	 * work that doesn't fit.
+	 * we use the passed in defer_ops and to finish up any work that
+	 * doesn't fit.  We need to reserve enough blocks to handle a
+	 * full btree split on either end of the refcount range.
 	 */
-	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, 0, 0, 0, &tp);
+	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate,
+			mp->m_refc_maxlevels * 2, 0, XFS_TRANS_RESERVE, &tp);
 	if (error)
 		return error;
 	cudp = xfs_trans_get_cud(tp, cuip);
diff --git a/fs/xfs/xfs_rmap_item.c b/fs/xfs/xfs_rmap_item.c
index bc1457c..e83a1da 100644
--- a/fs/xfs/xfs_rmap_item.c
+++ b/fs/xfs/xfs_rmap_item.c
@@ -23,6 +23,7 @@
 #include "xfs_log_format.h"
 #include "xfs_trans_resv.h"
 #include "xfs_bit.h"
+#include "xfs_shared.h"
 #include "xfs_mount.h"
 #include "xfs_defer.h"
 #include "xfs_trans.h"
@@ -473,7 +474,8 @@ xfs_rui_recover(
 		}
 	}
 
-	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, 0, 0, 0, &tp);
+	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate,
+			mp->m_rmap_maxlevels, 0, XFS_TRANS_RESERVE, &tp);
 	if (error)
 		return error;
 	rudp = xfs_trans_get_rud(tp, ruip);

Re: [PATCH] xfs: reserve blocks for refcount / rmap log item recovery

[Dave Chinner]

On Tue, Feb 20, 2018 at 07:38:54PM -0800, Darrick J. Wong wrote:
> From: Darrick J. Wong <darrick.wong@oracle.com>
> 
> During log recovery, the per-AG reservations aren't yet set up, so log
> recovery has to reserve enough blocks to handle all possible btree
> splits.
> 
> Reported-by: Dave Chinner <david@fromorbit.com>
> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>

This appears to have fixed all the random recovery failures I was
having through fstests. The logic seems sounds, too, so

Reviewed-by: Dave Chinner <dchinner@redhat.com>

-- 
Dave Chinner
david@fromorbit.com