From: Stephen Tweedie There's not much to be gained from making the resize group-add operation fully synchronous. It's already atomic by virtue of the journal, and a resize operation usually doesn't add just one group, but many. The real need is to sync things up at the end of the *entire* resize, not after each individual group-add; and by syncing each group separately, the whole resize operation is made much, much slower if we're running on a live filesystem. Signed-off-by: Stephen Tweedie Signed-off-by: Andrew Morton --- 25-akpm/fs/ext3/resize.c | 1 - 1 files changed, 1 deletion(-) diff -puN fs/ext3/resize.c~ext3-online-resize-make-group-add-asynchronous fs/ext3/resize.c --- 25/fs/ext3/resize.c~ext3-online-resize-make-group-add-asynchronous Thu Sep 30 17:33:51 2004 +++ 25-akpm/fs/ext3/resize.c Thu Sep 30 17:33:51 2004 @@ -872,7 +872,6 @@ int ext3_group_add(struct super_block *s exit_journal: unlock_super(sb); - handle->h_sync = 1; if ((err2 = ext3_journal_stop(handle)) && !err) err = err2; if (!err) { _