ocfs2: call ocfs2_journal_access_di() before ocfs2_journal_dirty() in ocfs2_write_end_nolock()

1.After we call ocfs2_journal_access_di() in ocfs2_write_begin(), jbd2_journal_restart() may also be called, in this function transaction A's t_updates-- and obtains a new transaction B. If jbd2_journal_commit_transaction() is happened to commit transaction A, when t_updates==0, it will continue to complete commit and unfile buffer.

So when jbd2_journal_dirty_metadata(), the handle is pointed a new transaction B, and the buffer head's journal head is already freed, jh->b_transaction == NULL, jh->b_next_transaction == NULL, it returns EINVAL, So it triggers the BUG_ON(status).

thread 1 jbd2
ocfs2_write_begin jbd2_journal_commit_transaction
ocfs2_write_begin_nolock
ocfs2_start_trans

     jbd2__journal_start(t_updates+1,
                        transaction A)
     ocfs2_journal_access_di
     ocfs2_write_cluster_by_desc
       ocfs2_mark_extent_written
         ocfs2_change_extent_flag
           ocfs2_split_extent
             ocfs2_extend_rotate_transaction
               jbd2_journal_restart
               (t_updates-1,transaction B) t_updates==0
                                         __jbd2_journal_refile_buffer
                                         (jh->b_transaction = NULL)

ocfs2_write_end
ocfs2_write_end_nolock
ocfs2_journal_dirty
jbd2_journal_dirty_metadata(bug)
ocfs2_commit_trans

2. In ext4, I found that: jbd2_journal_get_write_access() called by ext4_write_end.
ext4_write_begin

     ext4_journal_start
         __ext4_journal_start_sb
             ext4_journal_check_start
             jbd2__journal_start

ext4_write_end
ext4_mark_inode_dirty
ext4_reserve_inode_write
ext4_journal_get_write_access
jbd2_journal_get_write_access
ext4_mark_iloc_dirty
ext4_do_update_inode
ext4_handle_dirty_metadata
jbd2_journal_dirty_metadata

3.So I think we should put ocfs2_journal_access_di before ocfs2_journal_dirty in the ocfs2_write_end.
and it works well after my modification.

Signed-off-by vicky <[email protected]>

---
  fs/ocfs2/aops.c | 20 +++++++++-----------
  1 file changed, 9 insertions(+), 11 deletions(-)

fs/ocfs2/aops.c +9 -11

index d310d12..1f87c7c 100644
@@ -1818,16 +1818,6 @@ try_again:
  		if (ret)
  			goto out_commit;
  	}
-	/*
-	 * We don't want this to fail in ocfs2_write_end(), so do it
-	 * here.
-	 */
-	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), wc->w_di_bh,
-				      OCFS2_JOURNAL_ACCESS_WRITE);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_quota;
-	}
 
  	/*
  	 * Fill our page array first. That way we've grabbed enough so

@@ -2040,8 +2030,16 @@ out_write_size:
  	di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec);
  	di->i_mtime_nsec = di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
  	ocfs2_update_inode_fsync_trans(handle, inode, 1);
-	ocfs2_journal_dirty(handle, wc->w_di_bh);
+	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), wc->w_di_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret) {
+		copied = ret;
+		mlog_errno(ret);	
+		goto out;
+	}	
 
+	ocfs2_journal_dirty(handle, wc->w_di_bh);
+out:
  	ocfs2_commit_trans(osb, handle);
 
  	ocfs2_run_deallocs(osb, &wc->w_dealloc);

Wenfang,

You have to call the access function before modifying the buffer_head.
The access function takes care the case that the buffer_head is included in the previous transaction. In that case, the access function would make a copy of old contents(the right content for previous transaction) in the bh. If you modified the bh before calling access function, you may destroyed previous transaction.

For this problem, I'd suggest you to check all modification on this bh(for inode) and call access function before the first modification and make sure the transaction-restart safe.

thanks,
wengang

? 2014?06?09? 20:54, yangwenfang ??:

Quoted 90 lines

1.After we call ocfs2_journal_access_di() in ocfs2_write_begin(), jbd2_journal_restart() may also be called, in this function transaction A's t_updates-- and obtains a new transaction B. If jbd2_journal_commit_transaction() is happened to commit transaction A, when t_updates==0, it will continue to complete commit and unfile buffer.

So when jbd2_journal_dirty_metadata(), the handle is pointed a new transaction B, and the buffer head's journal head is already freed, jh->b_transaction == NULL, jh->b_next_transaction == NULL, it returns EINVAL, So it triggers the BUG_ON(status).

thread 1 jbd2
ocfs2_write_begin jbd2_journal_commit_transaction
ocfs2_write_begin_nolock
ocfs2_start_trans

     jbd2__journal_start(t_updates+1,
                        transaction A)
     ocfs2_journal_access_di
     ocfs2_write_cluster_by_desc
       ocfs2_mark_extent_written
         ocfs2_change_extent_flag
           ocfs2_split_extent
             ocfs2_extend_rotate_transaction
               jbd2_journal_restart
               (t_updates-1,transaction B) t_updates==0
                                         __jbd2_journal_refile_buffer
                                         (jh->b_transaction = NULL)

ocfs2_write_end
ocfs2_write_end_nolock
ocfs2_journal_dirty
jbd2_journal_dirty_metadata(bug)
ocfs2_commit_trans

2. In ext4, I found that: jbd2_journal_get_write_access() called by ext4_write_end.
ext4_write_begin

     ext4_journal_start
         __ext4_journal_start_sb
             ext4_journal_check_start
             jbd2__journal_start

ext4_write_end
ext4_mark_inode_dirty
ext4_reserve_inode_write
ext4_journal_get_write_access
jbd2_journal_get_write_access
ext4_mark_iloc_dirty
ext4_do_update_inode
ext4_handle_dirty_metadata
jbd2_journal_dirty_metadata

3.So I think we should put ocfs2_journal_access_di before ocfs2_journal_dirty in the ocfs2_write_end.
and it works well after my modification.

Signed-off-by vicky <[email protected]>

---
  fs/ocfs2/aops.c | 20 +++++++++-----------
  1 file changed, 9 insertions(+), 11 deletions(-)

fs/ocfs2/aops.c +3 -11

index d310d12..1f87c7c 100644
@@ -1818,16 +1818,6 @@ try_again:
          if (ret)
              goto out_commit;
      }
-    /*
-     * We don't want this to fail in ocfs2_write_end(), so do it
-     * here.
-     */
-    ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), wc->w_di_bh,
-                      OCFS2_JOURNAL_ACCESS_WRITE);
-    if (ret) {
-        mlog_errno(ret);
-        goto out_quota;
-    }
 
      /*
       * Fill our page array first. That way we've grabbed enough so

@@ -2040,8 +2030,16 @@ out_write_size:
      di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec);
      di->i_mtime_nsec = di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
      ocfs2_update_inode_fsync_trans(handle, inode, 1);
-    ocfs2_journal_dirty(handle, wc->w_di_bh);
+    ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), wc->w_di_bh,
+                      OCFS2_JOURNAL_ACCESS_WRITE);
+    if (ret) {

fs/ocfs2/aops.c out_write_size: · line 2045

Quoted 10 lines

fs/ocfs2/aops.c +6 -0

+        copied = ret;
+        mlog_errno(ret);   
+        goto out;
+    }   
 
+    ocfs2_journal_dirty(handle, wc->w_di_bh);
+out:
      ocfs2_commit_trans(osb, handle);
 
      ocfs2_run_deallocs(osb, &wc->w_dealloc);

.

fs/ocfs2/aops.c +3 -1

@@ -2040,8 +2030,16 @@ out_write_size:
 	di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec);
 	di->i_mtime_nsec = di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
 	ocfs2_update_inode_fsync_trans(handle, inode, 1);
-	ocfs2_journal_dirty(handle, wc->w_di_bh);
+	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), wc->w_di_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret) {