1# $NetBSD: README,v 1.3 1999/03/15 00:46:47 perseant Exp $ 2 3# @(#)README 8.1 (Berkeley) 6/11/93 4 5The file system is reasonably stable...I think. 6 7For details on the implementation, performance and why garbage 8collection always wins, see Dr. Margo Seltzer's thesis available for 9anonymous ftp from toe.cs.berkeley.edu, in the directory 10pub/personal/margo/thesis.ps.Z, or the January 1993 USENIX paper. 11 12---------- 13The disk is laid out in segments. The first segment starts 8K into the 14disk (the first 8K is used for boot information). Each segment is composed 15of the following: 16 17 An optional super block 18 One or more groups of: 19 segment summary 20 0 or more data blocks 21 0 or more inode blocks 22 23The segment summary and inode/data blocks start after the super block (if 24present), and grow toward the end of the segment. 25 26 _______________________________________________ 27 | | | | | 28 | summary | data/inode | summary | data/inode | 29 | block | blocks | block | blocks | ... 30 |_________|____________|_________|____________| 31 32The data/inode blocks following a summary block are described by the 33summary block. In order to permit the segment to be written in any order 34and in a forward direction only, a checksum is calculated across the 35blocks described by the summary. Additionally, the summary is checksummed 36and timestamped. Both of these are intended for recovery; the former is 37to make it easy to determine that it *is* a summary block and the latter 38is to make it easy to determine when recovery is finished for partially 39written segments. These checksums are also used by the cleaner. 40 41 Summary block (detail) 42 ________________ 43 | sum cksum | 44 | data cksum | 45 | next segment | 46 | timestamp | 47 | FINFO count | 48 | inode count | 49 | flags | 50 |______________| 51 | FINFO-1 | 0 or more file info structures, identifying the 52 | . | blocks in the segment. 53 | . | 54 | . | 55 | FINFO-N | 56 | inode-N | 57 | . | 58 | . | 59 | . | 0 or more inode daddr_t's, identifying the inode 60 | inode-1 | blocks in the segment. 61 |______________| 62 63Inode blocks are blocks of on-disk inodes in the same format as those in 64the FFS. However, spare[0] contains the inode number of the inode so we 65can find a particular inode on a page. They are packed page_size / 66sizeof(inode) to a block. Data blocks are exactly as in the FFS. Both 67inodes and data blocks move around the file system at will. 68 69The file system is described by a super-block which is replicated and 70occurs as the first block of the first and other segments. (The maximum 71number of super-blocks is MAXNUMSB). Each super-block maintains a list 72of the disk addresses of all the super-blocks. The super-block maintains 73a small amount of checkpoint information, essentially just enough to find 74the inode for the IFILE (fs->lfs_idaddr). 75 76The IFILE is visible in the file system, as inode number IFILE_INUM. It 77contains information shared between the kernel and various user processes. 78 79 Ifile (detail) 80 ________________ 81 | cleaner info | Cleaner information per file system. (Page 82 | | granularity.) 83 |______________| 84 | segment | Space available and last modified times per 85 | usage table | segment. (Page granularity.) 86 |______________| 87 | IFILE-1 | Per inode status information: current version #, 88 | . | if currently allocated, last access time and 89 | . | current disk address of containing inode block. 90 | . | If current disk address is LFS_UNUSED_DADDR, the 91 | IFILE-N | inode is not in use, and it's on the free list. 92 |______________| 93 94 95First Segment at Creation Time: 96_____________________________________________________________ 97| | | | | | | | 98| 8K pad | Super | summary | inode | ifile | root | l + f | 99| | block | | block | | dir | dir | 100|________|_______|_________|_______|_______|_______|_______| 101 ^ 102 Segment starts here. 103 104Some differences from the Sprite LFS implementation. 105 1061. The LFS implementation placed the ifile metadata and the super block 107 at fixed locations. This implementation replicates the super block 108 and puts each at a fixed location. The checkpoint data is divided into 109 two parts -- just enough information to find the IFILE is stored in 110 two of the super blocks, although it is not toggled between them as in 111 the Sprite implementation. (This was deliberate, to avoid a single 112 point of failure.) The remaining checkpoint information is treated as 113 a regular file, which means that the cleaner info, the segment usage 114 table and the ifile meta-data are stored in normal log segments. 115 (Tastes great, less filling...) 116 1172. The segment layout is radically different in Sprite; this implementation 118 uses something a lot like network framing, where data/inode blocks are 119 written asynchronously, and a checksum is used to validate any set of 120 summary and data/inode blocks. Sprite writes summary blocks synchronously 121 after the data/inode blocks have been written and the existence of the 122 summary block validates the data/inode blocks. This permits us to write 123 everything contiguously, even partial segments and their summaries, whereas 124 Sprite is forced to seek (from the end of the data inode to the summary 125 which lives at the end of the segment). Additionally, writing the summary 126 synchronously should cost about 1/2 a rotation per summary. 127 1283. Sprite LFS distinguishes between different types of blocks in the segment. 129 Other than inode blocks and data blocks, we don't. 130 1314. Sprite LFS traverses the IFILE looking for free blocks. We maintain a 132 free list threaded through the IFILE entries. 133 1345. The cleaner runs in user space, as opposed to kernel space. It shares 135 information with the kernel by reading/writing the IFILE and through 136 cleaner specific system calls. 137 138