On 06/25/2012 06:56 PM, Christoph Hellwig wrote: > On Sun, Jun 24, 2012 at 07:34:42PM +0800, Liu Yuan wrote: >> From: Liu Yuan <tailai.ly at taobao.com> >> >> We should also add the piggybacked ctime to the length > > Ooh. I think this needs some more fixes over sheepdog. For one > epoch_log_read should account for it directly when reading from the > file, that is we need to be able to read SD_MAX_NODES + the time_t. > > Probably better is to ignore the time in a plain epoch_log_read > and require a separate call for it, that uses lseek to seek to the > end of the file. And epoch_log_read really needs a comment about > this time piggybacking. > Hmm, yes, looks good to me. We need add a separate call to abstract out ctime handling, in order to exclude further improper use of epoch_log_read(). Thanks, Yuan |