xfs: store rtgroup information with a bmap intent

Make the bmap intent items take an active reference to the rtgroup
containing the space that is being mapped or unmapped.  We will need
this functionality once we start enabling rmap and reflink on the rt
volume.  Technically speaking we need it even for !rtgroups filesystems
to prevent the (dummy) rtgroup 0 from going away, even though this will
never happen.

As a bonus, we can rework the xfs_bmap_deferred_class tracepoint to use
the xfs_group object to figure out the type and group number, widen the
group block number field to fit 64-bit quantities, and get rid of the
now redundant opdev and rtblock fields.

Signed-off-by: Darrick J. Wong <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>

Author: Darrick J. Wong

fs/xfs/xfs_bmap_item.c

@@ -318,14 +318,16 @@ xfs_bmap_update_create_done(
 	return &budp->bud_item;
 }
 
-/* Take a passive ref to the AG containing the space we're mapping. */
+/* Take a passive ref to the group containing the space we're mapping. */
 static inline void
 xfs_bmap_update_get_group(
 	struct xfs_mount	*mp,
 	struct xfs_bmap_intent	*bi)
 {
+	enum xfs_group_type	type = XG_TYPE_AG;
+
 	if (xfs_ifork_is_realtime(bi->bi_owner, bi->bi_whichfork))
-		return;
+		type = XG_TYPE_RTG;
 
 	/*
 	 * Bump the intent count on behalf of the deferred rmap and refcount
@@ -335,7 +337,7 @@ xfs_bmap_update_get_group(
 	 * remains nonzero across the transaction roll.
 	 */
 	bi->bi_group = xfs_group_intent_get(mp, bi->bi_bmap.br_startblock,
-			XG_TYPE_AG);
+				type);
 }
 
 /* Add this deferred BUI to the transaction. */
@@ -344,8 +346,6 @@ xfs_bmap_defer_add(
 	struct xfs_trans	*tp,
 	struct xfs_bmap_intent	*bi)
 {
-	trace_xfs_bmap_defer(bi);
-
 	xfs_bmap_update_get_group(tp->t_mountp, bi);
 
 	/*
@@ -358,18 +358,9 @@ xfs_bmap_defer_add(
 	 */
 	if (bi->bi_type == XFS_BMAP_MAP)
 		bi->bi_owner->i_delayed_blks += bi->bi_bmap.br_blockcount;
-	xfs_defer_add(tp, &bi->bi_list, &xfs_bmap_update_defer_type);
-}
-
-/* Release a passive AG ref after finishing mapping work. */
-static inline void
-xfs_bmap_update_put_group(
-	struct xfs_bmap_intent	*bi)
-{
-	if (xfs_ifork_is_realtime(bi->bi_owner, bi->bi_whichfork))
-		return;
 
-	xfs_group_intent_put(bi->bi_group);
+	trace_xfs_bmap_defer(bi);
+	xfs_defer_add(tp, &bi->bi_list, &xfs_bmap_update_defer_type);
 }
 
 /* Cancel a deferred bmap update. */
@@ -382,7 +373,7 @@ xfs_bmap_update_cancel_item(
 	if (bi->bi_type == XFS_BMAP_MAP)
 		bi->bi_owner->i_delayed_blks -= bi->bi_bmap.br_blockcount;
 
-	xfs_bmap_update_put_group(bi);
+	xfs_group_intent_put(bi->bi_group);
 	kmem_cache_free(xfs_bmap_intent_cache, bi);
 }
 

fs/xfs/xfs_trace.h

@@ -3081,11 +3081,10 @@ DECLARE_EVENT_CLASS(xfs_bmap_deferred_class,
 	TP_ARGS(bi),
 	TP_STRUCT__entry(
 		__field(dev_t, dev)
-		__field(dev_t, opdev)
+		__field(enum xfs_group_type, type)
 		__field(xfs_agnumber_t, agno)
 		__field(xfs_ino_t, ino)
-		__field(xfs_agblock_t, agbno)
-		__field(xfs_fsblock_t, rtbno)
+		__field(unsigned long long, gbno)
 		__field(int, whichfork)
 		__field(xfs_fileoff_t, l_loff)
 		__field(xfs_filblks_t, l_len)
@@ -3094,20 +3093,25 @@ DECLARE_EVENT_CLASS(xfs_bmap_deferred_class,
 	),
 	TP_fast_assign(
 		struct xfs_inode	*ip = bi->bi_owner;
+		struct xfs_mount	*mp = ip->i_mount;
 
-		__entry->dev = ip->i_mount->m_super->s_dev;
-		if (xfs_ifork_is_realtime(ip, bi->bi_whichfork)) {
-			__entry->agno = 0;
-			__entry->agbno = 0;
-			__entry->rtbno = bi->bi_bmap.br_startblock;
-			__entry->opdev = ip->i_mount->m_rtdev_targp->bt_dev;
+		__entry->dev = mp->m_super->s_dev;
+		__entry->type = bi->bi_group->xg_type;
+		__entry->agno = bi->bi_group->xg_gno;
+		if (bi->bi_group->xg_type == XG_TYPE_RTG &&
+		    !xfs_has_rtgroups(mp)) {
+			/*
+			 * Legacy rt filesystems do not have allocation groups
+			 * ondisk.  We emulate this incore with one gigantic
+			 * rtgroup whose size can exceed a 32-bit block number.
+			 * For this tracepoint, we report group 0 and a 64-bit
+			 * group block number.
+			 */
+			__entry->gbno = bi->bi_bmap.br_startblock;
 		} else {
-			__entry->agno = XFS_FSB_TO_AGNO(ip->i_mount,
-						bi->bi_bmap.br_startblock);
-			__entry->agbno = XFS_FSB_TO_AGBNO(ip->i_mount,
-						bi->bi_bmap.br_startblock);
-			__entry->rtbno = 0;
-			__entry->opdev = __entry->dev;
+			__entry->gbno = xfs_fsb_to_gbno(mp,
+						bi->bi_bmap.br_startblock,
+						bi->bi_group->xg_type);
 		}
 		__entry->ino = ip->i_ino;
 		__entry->whichfork = bi->bi_whichfork;
@@ -3116,14 +3120,13 @@ DECLARE_EVENT_CLASS(xfs_bmap_deferred_class,
 		__entry->l_state = bi->bi_bmap.br_state;
 		__entry->op = bi->bi_type;
 	),
-	TP_printk("dev %d:%d op %s opdev %d:%d ino 0x%llx agno 0x%x agbno 0x%x rtbno 0x%llx %s fileoff 0x%llx fsbcount 0x%llx state %d",
+	TP_printk("dev %d:%d op %s ino 0x%llx %sno 0x%x gbno 0x%llx %s fileoff 0x%llx fsbcount 0x%llx state %d",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
 		  __print_symbolic(__entry->op, XFS_BMAP_INTENT_STRINGS),
-		  MAJOR(__entry->opdev), MINOR(__entry->opdev),
 		  __entry->ino,
+		  __print_symbolic(__entry->type, XG_TYPE_STRINGS),
 		  __entry->agno,
-		  __entry->agbno,
-		  __entry->rtbno,
+		  __entry->gbno,
 		  __print_symbolic(__entry->whichfork, XFS_WHICHFORK_STRINGS),
 		  __entry->l_loff,
 		  __entry->l_len,