xfs: don't log dirty ranges for ordered buffers
authorBrian Foster <bfoster@redhat.com>
Tue, 29 Aug 2017 17:08:38 +0000 (10:08 -0700)
committerDarrick J. Wong <darrick.wong@oracle.com>
Fri, 1 Sep 2017 17:55:30 +0000 (10:55 -0700)
Ordered buffers are attached to transactions and pushed through the
logging infrastructure just like normal buffers with the exception
that they are not actually written to the log. Therefore, we don't
need to log dirty ranges of ordered buffers. xfs_trans_log_buf() is
called on ordered buffers to set up all of the dirty state on the
transaction, buffer and log item and prepare the buffer for I/O.

Now that xfs_trans_dirty_buf() is available, call it from
xfs_trans_ordered_buf() so the latter is now mutually exclusive with
xfs_trans_log_buf(). This reflects the implementation of ordered
buffers and helps eliminate confusion over the need to log ranges of
ordered buffers just to set up internal log state.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
fs/xfs/libxfs/xfs_btree.c
fs/xfs/libxfs/xfs_ialloc.c
fs/xfs/xfs_trans_buf.c

index e0bcc4a59efdac646947f917f8c5d2a7713997fb..0b7905a4e27a86593735b869f957f8bb47e68e42 100644 (file)
@@ -4464,12 +4464,10 @@ xfs_btree_block_change_owner(
         * though, so everything is consistent in memory.
         */
        if (bp) {
-               if (cur->bc_tp) {
+               if (cur->bc_tp)
                        xfs_trans_ordered_buf(cur->bc_tp, bp);
-                       xfs_btree_log_block(cur, bp, XFS_BB_OWNER);
-               } else {
+               else
                        xfs_buf_delwri_queue(bp, bbcoi->buffer_list);
-               }
        } else {
                ASSERT(cur->bc_flags & XFS_BTREE_ROOT_IN_INODE);
                ASSERT(level == cur->bc_nlevels - 1);
index 1e0658a3f155a8c8e3514206f1b731305f804150..988bb3f3144664216bdf6880fd79838785a7b8db 100644 (file)
@@ -378,8 +378,6 @@ xfs_ialloc_inode_init(
                                 * transaction and pin the log appropriately.
                                 */
                                xfs_trans_ordered_buf(tp, fbuf);
-                               xfs_trans_log_buf(tp, fbuf, 0,
-                                                 BBTOB(fbuf->b_length) - 1);
                        }
                } else {
                        fbuf->b_flags |= XBF_DONE;
index 8c99813e5377d2789bd96272fcc1c6bcca1f4913..3089e80153693fe5092ce88bb948394eb5d714da 100644 (file)
@@ -560,16 +560,12 @@ xfs_trans_log_buf(
        struct xfs_buf_log_item *bip = bp->b_fspriv;
 
        ASSERT(first <= last && last < BBTOB(bp->b_length));
+       ASSERT(!(bip->bli_flags & XFS_BLI_ORDERED));
 
        xfs_trans_dirty_buf(tp, bp);
 
-       /*
-        * If we have an ordered buffer we are not logging any dirty range but
-        * it still needs to be marked dirty and that it has been logged.
-        */
        trace_xfs_trans_log_buf(bip);
-       if (!(bip->bli_flags & XFS_BLI_ORDERED))
-               xfs_buf_item_log(bip, first, last);
+       xfs_buf_item_log(bip, first, last);
 }
 
 
@@ -722,12 +718,11 @@ xfs_trans_inode_alloc_buf(
 }
 
 /*
- * Mark the buffer as ordered for this transaction. This means
- * that the contents of the buffer are not recorded in the transaction
- * but it is tracked in the AIL as though it was. This allows us
- * to record logical changes in transactions rather than the physical
- * changes we make to the buffer without changing writeback ordering
- * constraints of metadata buffers.
+ * Mark the buffer as ordered for this transaction. This means that the contents
+ * of the buffer are not recorded in the transaction but it is tracked in the
+ * AIL as though it was. This allows us to record logical changes in
+ * transactions rather than the physical changes we make to the buffer without
+ * changing writeback ordering constraints of metadata buffers.
  */
 void
 xfs_trans_ordered_buf(
@@ -739,9 +734,16 @@ xfs_trans_ordered_buf(
        ASSERT(bp->b_transp == tp);
        ASSERT(bip != NULL);
        ASSERT(atomic_read(&bip->bli_refcount) > 0);
+       ASSERT(!xfs_buf_item_dirty_format(bip));
 
        bip->bli_flags |= XFS_BLI_ORDERED;
        trace_xfs_buf_item_ordered(bip);
+
+       /*
+        * We don't log a dirty range of an ordered buffer but it still needs
+        * to be marked dirty and that it has been logged.
+        */
+       xfs_trans_dirty_buf(tp, bp);
 }
 
 /*