Btrfs: Avoid superfluous tree-log writeout

We allow two log transactions at a time, but use same flag
to mark dirty tree-log btree blocks. So we may flush dirty
blocks belonging to newer log transaction when committing a
log transaction. This patch fixes the issue by using two
flags to mark dirty tree-log btree blocks.

Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

5 files changed, 47 insertions, 31 deletions

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 02b6afbd7450..101940fab9b3 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -980,12 +980,12 @@ int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
 
         while (1) {
                 ret = find_first_extent_bit(&log_root_tree->dirty_log_pages,
-                                    0, &start, &end, EXTENT_DIRTY);
+                                0, &start, &end, EXTENT_DIRTY | EXTENT_NEW);
                 if (ret)
                         break;
 
-                clear_extent_dirty(&log_root_tree->dirty_log_pages,
-                                   start, end, GFP_NOFS);
+                clear_extent_bits(&log_root_tree->dirty_log_pages, start, end,
+                                  EXTENT_DIRTY | EXTENT_NEW, GFP_NOFS);
         }
         eb = fs_info->log_root_tree->node;
 
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 94627c4cc193..4a86508ce473 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -4919,8 +4919,16 @@ struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
         btrfs_set_buffer_uptodate(buf);
 
         if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
-                set_extent_dirty(&root->dirty_log_pages, buf->start,
-                         buf->start + buf->len - 1, GFP_NOFS);
+                /*
+                 * we allow two log transactions at a time, use different
+                 * EXENT bit to differentiate dirty pages.
+                 */
+                if (root->log_transid % 2 == 0)
+                        set_extent_dirty(&root->dirty_log_pages, buf->start,
+                                        buf->start + buf->len - 1, GFP_NOFS);
+                else
+                        set_extent_new(&root->dirty_log_pages, buf->start,
+                                        buf->start + buf->len - 1, GFP_NOFS);
         } else {
                 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
                          buf->start + buf->len - 1, GFP_NOFS);
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index c207e8c32c9b..b7b22c344b66 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -354,7 +354,7 @@ int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
  * those extents are sent to disk but does not wait on them
  */
 int btrfs_write_marked_extents(struct btrfs_root *root,
-                               struct extent_io_tree *dirty_pages)
+                               struct extent_io_tree *dirty_pages, int mark)
 {
         int ret;
         int err = 0;
@@ -367,7 +367,7 @@ int btrfs_write_marked_extents(struct btrfs_root *root,
 
         while (1) {
                 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
-                                            EXTENT_DIRTY);
+                                            mark);
                 if (ret)
                         break;
                 while (start <= end) {
@@ -413,7 +413,7 @@ int btrfs_write_marked_extents(struct btrfs_root *root,
  * on all the pages and clear them from the dirty pages state tree
  */
 int btrfs_wait_marked_extents(struct btrfs_root *root,
-                              struct extent_io_tree *dirty_pages)
+                              struct extent_io_tree *dirty_pages, int mark)
 {
         int ret;
         int err = 0;
@@ -425,12 +425,12 @@ int btrfs_wait_marked_extents(struct btrfs_root *root,
         unsigned long index;
 
         while (1) {
-                ret = find_first_extent_bit(dirty_pages, 0, &start, &end,
-                                            EXTENT_DIRTY);
+                ret = find_first_extent_bit(dirty_pages, start, &start, &end,
+                                            mark);
                 if (ret)
                         break;
 
-                clear_extent_dirty(dirty_pages, start, end, GFP_NOFS);
+                clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
                 while (start <= end) {
                         index = start >> PAGE_CACHE_SHIFT;
                         start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
@@ -460,13 +460,13 @@ int btrfs_wait_marked_extents(struct btrfs_root *root,
  * those extents are on disk for transaction or log commit
  */
 int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
-                                        struct extent_io_tree *dirty_pages)
+                                struct extent_io_tree *dirty_pages, int mark)
 {
         int ret;
         int ret2;
 
-        ret = btrfs_write_marked_extents(root, dirty_pages);
-        ret2 = btrfs_wait_marked_extents(root, dirty_pages);
+        ret = btrfs_write_marked_extents(root, dirty_pages, mark);
+        ret2 = btrfs_wait_marked_extents(root, dirty_pages, mark);
         return ret || ret2;
 }
 
@@ -479,7 +479,8 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
                 return filemap_write_and_wait(btree_inode->i_mapping);
         }
         return btrfs_write_and_wait_marked_extents(root,
-                                           &trans->transaction->dirty_pages);
+                                           &trans->transaction->dirty_pages,
+                                           EXTENT_DIRTY);
 }
 
 /*
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index d4e3e7a6938c..93c7ccb33118 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -107,10 +107,10 @@ void btrfs_throttle(struct btrfs_root *root);
 int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans,
                                 struct btrfs_root *root);
 int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
-                                        struct extent_io_tree *dirty_pages);
+                                struct extent_io_tree *dirty_pages, int mark);
 int btrfs_write_marked_extents(struct btrfs_root *root,
-                                        struct extent_io_tree *dirty_pages);
+                                struct extent_io_tree *dirty_pages, int mark);
 int btrfs_wait_marked_extents(struct btrfs_root *root,
-                                        struct extent_io_tree *dirty_pages);
+                                struct extent_io_tree *dirty_pages, int mark);
 int btrfs_transaction_in_commit(struct btrfs_fs_info *info);
 #endif
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 741666a7676a..31da0002e78b 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -1977,10 +1977,11 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 {
         int index1;
         int index2;
+        int mark;
         int ret;
         struct btrfs_root *log = root->log_root;
         struct btrfs_root *log_root_tree = root->fs_info->log_root_tree;
-        u64 log_transid = 0;
+        unsigned long log_transid = 0;
 
         mutex_lock(&root->log_mutex);
         index1 = root->log_transid % 2;
@@ -2014,24 +2015,29 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
                 goto out;
         }
 
+        log_transid = root->log_transid;
+        if (log_transid % 2 == 0)
+                mark = EXTENT_DIRTY;
+        else
+                mark = EXTENT_NEW;
+
         /* we start IO on  all the marked extents here, but we don't actually
          * wait for them until later.
          */
-        ret = btrfs_write_marked_extents(log, &log->dirty_log_pages);
+        ret = btrfs_write_marked_extents(log, &log->dirty_log_pages, mark);
         BUG_ON(ret);
 
         btrfs_set_root_node(&log->root_item, log->node);
 
         root->log_batch = 0;
-        log_transid = root->log_transid;
         root->log_transid++;
         log->log_transid = root->log_transid;
         root->log_start_pid = 0;
         smp_mb();
         /*
-         * log tree has been flushed to disk, new modifications of
-         * the log will be written to new positions. so it's safe to
-         * allow log writers to go in.
+         * IO has been started, blocks of the log tree have WRITTEN flag set
+         * in their headers. new modifications of the log will be written to
+         * new positions. so it's safe to allow log writers to go in.
          */
         mutex_unlock(&root->log_mutex);
 
@@ -2052,7 +2058,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 
         index2 = log_root_tree->log_transid % 2;
         if (atomic_read(&log_root_tree->log_commit[index2])) {
-                btrfs_wait_marked_extents(log, &log->dirty_log_pages);
+                btrfs_wait_marked_extents(log, &log->dirty_log_pages, mark);
                 wait_log_commit(trans, log_root_tree,
                                 log_root_tree->log_transid);
                 mutex_unlock(&log_root_tree->log_mutex);
@@ -2072,16 +2078,17 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
          * check the full commit flag again
          */
         if (root->fs_info->last_trans_log_full_commit == trans->transid) {
-                btrfs_wait_marked_extents(log, &log->dirty_log_pages);
+                btrfs_wait_marked_extents(log, &log->dirty_log_pages, mark);
                 mutex_unlock(&log_root_tree->log_mutex);
                 ret = -EAGAIN;
                 goto out_wake_log_root;
         }
 
         ret = btrfs_write_and_wait_marked_extents(log_root_tree,
-                                &log_root_tree->dirty_log_pages);
+                                &log_root_tree->dirty_log_pages,
+                                EXTENT_DIRTY | EXTENT_NEW);
         BUG_ON(ret);
-        btrfs_wait_marked_extents(log, &log->dirty_log_pages);
+        btrfs_wait_marked_extents(log, &log->dirty_log_pages, mark);
 
         btrfs_set_super_log_root(&root->fs_info->super_for_commit,
                                 log_root_tree->node->start);
@@ -2147,12 +2154,12 @@ int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root)
 
         while (1) {
                 ret = find_first_extent_bit(&log->dirty_log_pages,
-                                    0, &start, &end, EXTENT_DIRTY);
+                                0, &start, &end, EXTENT_DIRTY | EXTENT_NEW);
                 if (ret)
                         break;
 
-                clear_extent_dirty(&log->dirty_log_pages,
-                                   start, end, GFP_NOFS);
+                clear_extent_bits(&log->dirty_log_pages, start, end,
+                                  EXTENT_DIRTY | EXTENT_NEW, GFP_NOFS);
         }
 
         if (log->log_transid > 0) {