xfs: add a free space extent change reservation

Lots of the transaction reservation code reserves space for an
extent allocation. It is inconsistently implemented, and many of
them get it wrong. Introduce a new function to calculate the log
space reservation for adding or removing an extent from the free
space btrees.

This function reserves space for logging the AGF, the AGFL and the
free space btrees, avoiding the need to account for them seperately
in every reservation that manipulates free space.

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

fs/xfs/libxfs/xfs_trans_resv.c

@@ -79,6 +79,23 @@ xfs_allocfree_log_count(
 	return blocks;
 }
 
+/*
+ * Log reservation required to add or remove a single extent to the free space
+ * btrees.  This requires modifying:
+ *
+ * the agf header: 1 sector
+ * the agfl header: 1 sector
+ * the allocation btrees: 2 trees * (max depth - 1) * block size
+ */
+uint
+xfs_allocfree_extent_res(
+	struct xfs_mount *mp)
+{
+	return xfs_calc_buf_res(2, mp->m_sb.sb_sectsize) +
+	       xfs_calc_buf_res(xfs_allocfree_log_count(mp, 1),
+				XFS_FSB_TO_B(mp, 1));
+}
+
 /*
  * Logging inodes is really tricksy. They are logged in memory format,
  * which means that what we write into the log doesn't directly translate into