Updates for 4.14: - Write unmount record for a ro mount to avoid unnecessary log replay - Clean up orphaned inodes when mounting fs readonly - Resubmit inode log items when buffer writeback fails to avoid umount hang - Fix log recovery corruption problems when log headers wrap around the end - Avoid infinite loop searching for free inodes when inode counters are wrong - Evict inodes involved with log redo so that we don't leak them later - Fix a potential race between reclaim and inode cluster freeing - Refactor the inode joining code w.r.t. transaction rolling & deferred ops - Fix a bug where the log doesn't properly deal with dirty buffers that are about to become ordered buffers - Fix the extent swap code to deal with making dirty buffers ordered properly - Consolidate page fault handlers - Refactor the incore extent manipulation functions to use the iext abstractions instead of directly modifying with extent data - Disable crashy chattr +/-x until we fix it - Don't allow us to set S_DAX for v2 inodes - Various cleanups - Clarify some documentation - Fix a problem where fsync and a log commit race to send the disk a flush command, resulting in a small window where power fail data loss could occur - Simplify some rmap operations in the fcollapse code