Shortening fsck Time on ext2

Shortening fsck time on ext2

Val Henson, Intel, Theodore Y. T'so, IBM, Zach Brown, Oracle, Arjan van de Ven, Intel

Ext2 is fast, simple, robust, and fun to hack on.  However, it has fallen out of favor for one major reason: if an ext2 file system is not cleanly unmounted, such as in the event of kernel crash or power loss, it must be repaired using fsck, which takes minutes or hours to complete, during which time the file system is unavailable.  In this paper, we describe some techniques for reducing the average fsck time on ext2 file systems.  First, we avoid running fsck in some cases by adding a file system wide dirty bit indicating whether the file system was being actively modified at the time it crashed.  The performance of ext2 with this change is close to that of plain ext2, and quite a bit faster than ext3.  Second, we propose a technique called linked writes which uses dependent writes and a list of dirty inodes to allow recovery of an active file system by only repairing the dirty inodes and avoiding a full file system check.