orphan.rst (2450B)
1.. SPDX-License-Identifier: GPL-2.0 2 3Orphan file 4----------- 5 6In unix there can inodes that are unlinked from directory hierarchy but that 7are still alive because they are open. In case of crash the filesystem has to 8clean up these inodes as otherwise they (and the blocks referenced from them) 9would leak. Similarly if we truncate or extend the file, we need not be able 10to perform the operation in a single journalling transaction. In such case we 11track the inode as orphan so that in case of crash extra blocks allocated to 12the file get truncated. 13 14Traditionally ext4 tracks orphan inodes in a form of single linked list where 15superblock contains the inode number of the last orphan inode (s_last_orphan 16field) and then each inode contains inode number of the previously orphaned 17inode (we overload i_dtime inode field for this). However this filesystem 18global single linked list is a scalability bottleneck for workloads that result 19in heavy creation of orphan inodes. When orphan file feature 20(COMPAT_ORPHAN_FILE) is enabled, the filesystem has a special inode 21(referenced from the superblock through s_orphan_file_inum) with several 22blocks. Each of these blocks has a structure: 23 24============= ================ =============== =============================== 25Offset Type Name Description 26============= ================ =============== =============================== 270x0 Array of Orphan inode Each __le32 entry is either 28 __le32 entries entries empty (0) or it contains 29 inode number of an orphan 30 inode. 31blocksize-8 __le32 ob_magic Magic value stored in orphan 32 block tail (0x0b10ca04) 33blocksize-4 __le32 ob_checksum Checksum of the orphan block. 34============= ================ =============== =============================== 35 36When a filesystem with orphan file feature is writeably mounted, we set 37RO_COMPAT_ORPHAN_PRESENT feature in the superblock to indicate there may 38be valid orphan entries. In case we see this feature when mounting the 39filesystem, we read the whole orphan file and process all orphan inodes found 40there as usual. When cleanly unmounting the filesystem we remove the 41RO_COMPAT_ORPHAN_PRESENT feature to avoid unnecessary scanning of the orphan 42file and also make the filesystem fully compatible with older kernels.