ocfs2: do not change i_size in write_end for direct io
authorRyan Ding <ryan.ding@oracle.com>
Fri, 25 Mar 2016 21:21:01 +0000 (14:21 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 25 Mar 2016 23:37:42 +0000 (16:37 -0700)
To support direct io in ocfs2_write_begin_nolock & ocfs2_write_end_nolock.

Append direct io do not change i_size in get block phase.  It only move
to orphan when starting write.  After data is written to disk, it will
delete itself from orphan and update i_size.  So skip i_size change
section in write_begin for direct io.

And when there is no extents alloc, no meta data changes needed for
direct io (since write_begin start trans for 2 reason: alloc extents &
change i_size.  Now none of them needed).  So we can skip start trans
procedure.

Signed-off-by: Ryan Ding <ryan.ding@oracle.com>
Reviewed-by: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <joseph.qi@huawei.com>
Cc: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/ocfs2/aops.c

index 100f7a8..4ad7745 100644 (file)
@@ -2043,8 +2043,10 @@ static int ocfs2_expand_nonsparse_inode(struct inode *inode,
        if (ret)
                mlog_errno(ret);
 
-       wc->w_first_new_cpos =
-               ocfs2_clusters_for_bytes(inode->i_sb, i_size_read(inode));
+       /* There is no wc if this is call from direct. */
+       if (wc)
+               wc->w_first_new_cpos =
+                       ocfs2_clusters_for_bytes(inode->i_sb, i_size_read(inode));
 
        return ret;
 }
@@ -2135,14 +2137,17 @@ try_again:
                }
        }
 
-       if (ocfs2_sparse_alloc(osb))
-               ret = ocfs2_zero_tail(inode, di_bh, pos);
-       else
-               ret = ocfs2_expand_nonsparse_inode(inode, di_bh, pos, len,
-                                                  wc);
-       if (ret) {
-               mlog_errno(ret);
-               goto out;
+       /* Direct io change i_size late, should not zero tail here. */
+       if (type != OCFS2_WRITE_DIRECT) {
+               if (ocfs2_sparse_alloc(osb))
+                       ret = ocfs2_zero_tail(inode, di_bh, pos);
+               else
+                       ret = ocfs2_expand_nonsparse_inode(inode, di_bh, pos,
+                                                          len, wc);
+               if (ret) {
+                       mlog_errno(ret);
+                       goto out;
+               }
        }
 
        ret = ocfs2_check_range_for_refcount(inode, pos, len);
@@ -2203,8 +2208,9 @@ try_again:
 
                credits = ocfs2_calc_extend_credits(inode->i_sb,
                                                    &di->id2.i_list);
-
-       }
+       } else if (type == OCFS2_WRITE_DIRECT)
+               /* direct write needs not to start trans if no extents alloc. */
+               goto success;
 
        /*
         * We have to zero sparse allocated clusters, unwritten extent clusters,
@@ -2402,12 +2408,14 @@ int ocfs2_write_end_nolock(struct address_space *mapping,
        handle_t *handle = wc->w_handle;
        struct page *tmppage;
 
-       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;
+       if (handle) {
+               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;
+               }
        }
 
        if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
@@ -2450,25 +2458,29 @@ int ocfs2_write_end_nolock(struct address_space *mapping,
                }
 
                if (page_has_buffers(tmppage)) {
-                       if (ocfs2_should_order_data(inode))
-                               ocfs2_jbd2_file_inode(wc->w_handle, inode);
+                       if (handle && ocfs2_should_order_data(inode))
+                               ocfs2_jbd2_file_inode(handle, inode);
                        block_commit_write(tmppage, from, to);
                }
        }
 
 out_write_size:
-       pos += copied;
-       if (pos > i_size_read(inode)) {
-               i_size_write(inode, pos);
-               mark_inode_dirty(inode);
-       }
-       inode->i_blocks = ocfs2_inode_sector_count(inode);
-       di->i_size = cpu_to_le64((u64)i_size_read(inode));
-       inode->i_mtime = inode->i_ctime = CURRENT_TIME;
-       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);
+       /* Direct io do not update i_size here. */
+       if (wc->w_type != OCFS2_WRITE_DIRECT) {
+               pos += copied;
+               if (pos > i_size_read(inode)) {
+                       i_size_write(inode, pos);
+                       mark_inode_dirty(inode);
+               }
+               inode->i_blocks = ocfs2_inode_sector_count(inode);
+               di->i_size = cpu_to_le64((u64)i_size_read(inode));
+               inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+               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);
+       }
+       if (handle)
+               ocfs2_journal_dirty(handle, wc->w_di_bh);
 
 out:
        /* unlock pages before dealloc since it needs acquiring j_trans_barrier
@@ -2478,7 +2490,8 @@ out:
         */
        ocfs2_unlock_pages(wc);
 
-       ocfs2_commit_trans(osb, handle);
+       if (handle)
+               ocfs2_commit_trans(osb, handle);
 
        ocfs2_run_deallocs(osb, &wc->w_dealloc);