Hello everyone, I've hit a few bumps with the addres space operations, and I suspect the other journaled filesystems will have similar issues. generic_commit_write changes the inode->i_size, but doesn't mark the inode dirty. Since reiserfs logs inodes instead of calling mark_inode_dirty, it is likely the i_size change will never get to disk. ext2 calls mark_inode_dirty enough that it is probably safe, but there might be paths were the inode won't get dirtied. It would be faster if get_block could know the new file size. In reiserfs_get_block, I'm already logging the inode when new blocks are appended to the file. If I could update i_size at the same time, I would not have to log or dirty it in commit_write. This would be a big gain for reiserfs, as inode updates are somewhat expensive. -chris
