嵌入式 Linux 中的文件系统 : 设计思想和工作原理

Transcription

1 嵌入式 Linux 中的文件系统 : 设计思想和工作原理 刘肖凡博士副教授东南大学计算机学院

2 关于我 刘肖凡 东南大学计算机科学与工程学院副教授 计算机工程系副主任 香港理工大学电子及资讯工程学系学士 博士 3 年教授 操作系统原理 课程经验 研究方向 : 大数据 复杂网络 江苏省软件工程标准化委员会委员 中国计算机协会青年科技论坛南京分论坛学术委员

3 内容提要 文件系统概念回顾 主流嵌入式 Linux 文件系统 如何选择合适的文件系统

4 内容提要 文件系统概念回顾 主流嵌入式 Linux 文件系统 如何选择合适的文件系统

5 文件系统概念回顾 文件 元数据 文件系统安装 存取方法 小结

6 文件系统概念回顾 文件 元数据 文件系统安装 存取方法 小结

7 文件 -File Contiguous logical address space Types: Data numeric character binary Program source file, executable file

8 文件系统概念回顾 文件 元数据 文件系统安装 存取方法 小结

9 元数据 -Metadata Name Identifier unique tag (number) identifies file within file system Type needed for systems that support different types Location pointer to file location on device Size current file size Protection controls who can do reading, writing, executing Time, date, and user identification data for protection, security, and usage monitoring Many variations, including extended file attributes such as file checksum Not all included in embedded file systems

10 元数据 -Metadata (Cont.) Per-file File Control Block (FCB) contains many details about the file inode stucture in UNIX stuct inode or vnode in Linux Master File Table using relational DB structures in NFTS

11 Directory Structure Directory structure organizes the files Names and inode numbers, master file table Basic data structure includes linear list, hash table, etc.. Each volume containing file system also tracks that file system s info in device directory Boot control block contains info needed by system to boot OS from that volume Needed if volume contains OS, usually first block of volume Volume control block (superblock, master file table) contains volume details Total # of blocks, # of free blocks, block size, free block pointers or array

12 A Typical File-system Org.

13 In-Memory FS Structures

14 Operations on Files Directory level Search for a file - returns a file handle Create a file Delete a file List a directory Rename a file Traverse the file system Content level Write to buffers first Read - copied to user process memory address Reposition within file Truncate

15 文件系统概念回顾 文件 元数据 文件系统安装 存取方法 小结

16 文件系统安装 -Mounting A file system must be mounted before it can be accessed A unmounted file system (i.e., Fig (b)) is mounted at a mount point

17 文件系统概念回顾 文件 元数据 文件系统安装 存取方法 小结

18 存取方法 -Accessing Files Character-based device Sequential device Block device

19 Sequential Device

20 Block Device Disk provides in-place rewrite and random access I/O transfers performed in blocks of sectors (usually 512 bytes)

21 Other Access Methods Can be built on top of base methods General involve creation of an index for the file Keep index in memory for fast determination of location of data to be operated on (consider UPC code plus record of data about that item) If too large, index (in memory) of the index (on disk)

22 Flash Memory*

23 文件系统概念回顾 文件 元数据 文件系统安装 存取方法 小结

24 File-System Structure File structure Logical storage unit Collection of related information File system resides on secondary storage Provided user interface to storage, mapping logical to physical Provides efficient and convenient access to disk by allowing data to be stored, located retrieved easily File system organized into layers

25 Layered File System

26 File System Layers Device drivers manage I/O devices at the I/O control layer Given commands like read drive1, cylinder 72, track 2, sector 10, into memory location 1060 outputs low-level hardware specific commands to hardware controller Basic file system given command like retrieve block 123 translates to device driver Also manages memory buffers and caches (allocation, freeing, replacement) Buffers hold data in transit Caches hold frequently used data

27 File System Layers (Cont.) File organization module understands files, logical address, and physical blocks Translates logical block # to physical block # Manages free space, disk allocation Logical file system manages metadata information File control lock Directory management Protection

28 Linux Kernel and File System A file system must be present for the kernel to start successfully Kernel and file system can change independently Can be an in memory file system, network file system Can be attached to the kernel image loaded into memory The Linux kernel, after starting, will mount the file system and execute some program While they may be packaged together, the root file system is a separate entity from the kernel Root file system Mounted at / A module separated from the kernel

29 内容提要 文件系统概念回顾 主流嵌入式 Linux 文件系统 如何选择合适的文件系统

30 主流嵌入式 Linux 文件系统 二级储存器 适合闪存使用的文件系统 针对 MTD 的文件系统 自带 FTL 的文件系统 其他嵌入式文件系统

31 主流嵌入式 Linux 文件系统 二级储存器 适合闪存使用的文件系统 针对 MTD 的文件系统 自带 FTL 的文件系统 其他嵌入式文件系统

32 二级储存器 Magnetic tape Magnetic disk Flash memory suitable for embedded system

33 Flash Memory Flash memory is an electronic non-volatile computer storage medium that can be electrically erased and reprogrammed. Two main types of flash memory : NAND and NOR The internal characteristics of the individual flash memory cells exhibit characteristics similar to those of the corresponding gates. Where as EPROMs had to be completely erased before being rewritten, NAND type flash memory may be written and read in blocks (or pages) which are generally much smaller than the entire device. NOR type flash allows a single machine word (byte) to be written to an erased location or read independently.

34 NAND vs. NOR NAND: primarily used in main memory, memory cards, USB flash drives, solid-state drives, etc. NOR: allows true random access and therefore direct code execution.. NAND NOR Main Application File Storage Code execution Storage capacity High Low Cost per bit Better. Active Power Better. Standby Power. Better Write Speed Good. Read Speed. Good

35 Disadvantage of Flash Memory One significant disadvantage of flash memory is the finite amount of read/write cycles in a specific block. On low level, they require special handling such as wear leveling and different error detection and correction algorithms. Typically a device such as solid-state disk handles such operations internally and therefore a regular file system can be used. However, for certain specialized installations (embedded systems, industrial applications) a file system optimized for plain flash memory is needed.

36 主流嵌入式 Linux 文件系统 二级储存器 适合闪存使用的文件系统 针对 MTD 的文件系统 自带 FTL 的文件系统 其他嵌入式文件系统

37 Utilizing Raw Flash Device Both abstraction layers for utilizing raw flash devices MTD (Memory Technology Devices) /dev/mtdblockx or /dev/mtd0 Does not handle wearing and bad blocks, etc. FTL (Flash Translation Layer) Handle wearing and bad blocks, etc. Looks like block device

38 MTD vs. Block device MTD device Consists of eraseblocks Eraseblocks are larger (typically 128KB) Maintains 3 main operations: read from eraseblock, write to eraseblock, and erase eraseblock Bad eraseblocks are not hidden and should be dealt with in software Eraseblocks wear-out and become bad and unusable after about 10 3 (for MLC NAND) (NOR, SLC NAND) erase cycles Block device Consists of sectors Sectors are small (512, 1024 bytes) Maintains 2 main operations: read sector and write sector Bad sectors are re-mapped and hidden by hardware (at least in modern LBA hard drives); in case of FTL devices it is the responsibility of FTL to provide this Sectors are devoid of the wear-out property (in FTL devices it is the responsibility of FTL to provide this)

39 FS in Embedded Linux Block devices ext2 Very stable, easy to work with, widely supported, keeps all permissions but, not very space efficient cramfs Produces a small file system tradeoff: read-only with minimal permissions squashfs More metadata and larger file systems, great compression results in small file system, but performance hit romfs Small kernel module, but lacks compression Flash JFFS2 Stores all metadata, high capacity performance lacking on mount times and writes (under certain circumstances) YAFFS2 Handles particularities of NAND flash performance also lacking under certain circumstances In Memory initramfs Complete support for permissions and file ownership, however stored in memory, so changes aren t persistent

40 Proprietary Flash FS Microsoft FAT FATX FAT+ exfat TFAT FFS2 HPFS Sansumg F2FS RFS IBM JFS Palm NVFS Network Appliance WAFL

41 Proprietary Flash FS (Cont.) CASL is a filesystem designed by Nimble Storage that uses solid-state devices to cache traditional hard disk drives. ETFS - Embedded Transactional File System. Designed primarily for NAND devices by QNX Software Systems. OneFS - OneFS is a file system utilized by Isilon. It supports selective placement of meta-data directly onto flash SSD. Segger Microcontroller Systems emfile - File system for deeply embedded applications which supports both NAND and NOR flashes. Wear leveling, fast read and write, and very low RAM usage. SafeFLASH - HCC-Embedded - Fail-safe file system that supports NAND and NOR flash types with integrated wear-leveling and bad-block handling. Unison RTOS - Fsys-Nand/Nor small footprint low cost flash file system for embedded systems ZFS - Allows placing write-ahead log (ZIL) on flash, and using flash as a second-level read cache (L2ARC) OTFS - Used in BOS

42 主流嵌入式 Linux 文件系统 二级储存器 适合闪存使用的文件系统 针对 MTD 的文件系统 自带 FTL 的文件系统 其他嵌入式文件系统

43 Log-structured FS Different from traditional in-place rewrite file systems Log-structured file systems write in empty space. Taking WAFL as an example Note: WAFL is originally designed as a embedded system file system but now used as a network file system.

44 The WAFL File Layout

45 Wear Leveling In MTD

46 JFFS/JFFS2 Description Original log-structured Linux file system for NOR flash media. Handles wear leveling When to Use Capacity and Limitations NOR and NAND devices 2 32 GB, 2 32 files, Page size from 2 12 to 2 18 ; Complete POSIX meta data; Mounts slowly (improved lately) At capacity, writes can be slow Compresses data during creation and subsequent writes How to Use $ mkfs.jffs2 -o../<bsp_name>-flash.jffs2 -e (Can specify erase block size) mount the /dev/mtdblock device as the root device. Home Page More Info

47 YAFFS1/YAFFS2 Description Works, in principle, much like JFFS2 But designed specifically for NAND flash devices. Tag blocks with monotonically increasing seq. number When to Use Capacity and Limitations NAND flash devices 2 32 GB, 2 32 files Complete POSIX meta data Data integrity Mount very fast (through checkpointing) How to Use File systems created using user space tool, much like JFFS2. The resulting file can then be written directly to a flash partition. Migrating functionalities to JFFS2 Home Page More Info

48 LogFS Description LogFS is a Linux log-structured and scalable flash file system intended for use on large devices of flash memory. When to Use Capacity and Limitations How to Use Home Page More Info large devices of flash memory Fast compression Mount very fast (stores the inode tree on the drive) Quicker access to large files File systems created using user space tool, much like JFFS2.

49 UBIFS Description A successor to JFFS2, and competitor to LogFS. Supports write caching UBIFS errs on the pessimistic side of free space calculation, When to Use Large NAND FLASH devices On unsorted block images (UBI) device, which is on top of a MTD Capacity and Limitations How to Use Home Page More Info Complete POSIX meta data On-the-fly compression Recoverability and power fail tolerance Faster mounting (in-flash directory structure) Quicker access to large files Improved write speeds $ mount -o no_chk_data_crc /dev/ubi0_0 /mnt/ubifs The resulting file can then be written directly to a flash partition.

50 主流嵌入式 Linux 文件系统 二级储存器 适合闪存使用的文件系统 针对 MTD 的文件系统 自带 FTL 的文件系统 其他嵌入式文件系统

51 Flash Translation Layer A flash translation layer is used to adapt a fully functional file system to the constraints and restrictions imposed by flash memory devices. Wear-leveling Error detection Generally shipped with SSD drives.

52 TrueFFS and ExtremeFFS Both FTL for SSDs, implementing error correction, bad block re-mapping and wear levelling. Not a file system, no file system interface but only disk interface. ExtremeFFS being successor of TrueFFS. Improved random write performance (by a factor of 100). On top of FTL, all file systems could be used. ext*/f2fs/fat Consider your own SSD drive.

53 主流嵌入式 Linux 文件系统 二级储存器 适合闪存使用的文件系统 针对 MTD 的文件系统 自带 FTL 的文件系统 其他嵌入式文件系统

54 initramfs Linux default file system Integral part of 2.6 Linux kernel boot A file system that sits on top of the kernel s inode cache Looks for initramfs before using traditional booting method Can use as real filesystem Can be replaced by other file system on boot time Used in no write scenario

55 initramfs How to create Part of the kernel build process As a compressed cpio archive $ cd <rfs-directory> $ find. cpio -o -H newc gzip >../initramfs_data.cpio.gz Point to a directory Make CONFIG_INITRAMFS_SOURCE a directory name Use specification file Make CONFIG_INITRAMFS_SOURCE a file name that specifies what files\devices to create with what ownership permissions More Information

56 cramfs Description Compressed ROM File system Read only file system widely used in the embedded space Data stored in compressed format (zlib) When to Use Low-memory systems Ensures RFS integrity Metadata not important (doesn t store full information) Capacity and Limitations How to Use Home Page More Info 256 MB, 2 16 files Does not store all permissions, all files owned by root. No timestamps stored (inode overhead is just 12 bytes!) $ mkcramfs -m dev.cramfs.txt <rfs_dir> rootfs.cramfs

57 squashfs Description Read only file system that includes several improvements over cramfs, notably in compression and metadata storage. Adjustable block sizes allow a user to create filesystems that compress better. When to Use Low-memory systems Need control over the endianness Capacity and Limitations How to Use Home Page More Info 2 32 GB, 2 32 files, Page size from 2 12 to 2 18 Read-only $ mksquashfs $RFS./squashfs-rfs/rfs -nopad -all-root The resulting file can then be written directly to a flash partition

58 romfs Description Minimum filesystem, very small kernel module. When to Use Trying to make as compact a kernel as possible Initial RAM disks Capacity and Limitations How to Use Home Page More Info Read-only No compression $ genromfs -f./romfs-rfs/rfs -d $RFS Create filesystem with mkromfs utility Creating device nodes particularly interesting create a file starting with device node information <kernel>/documentation/filesystems

59 How These Stack Up Results (in Bytes) 305,376 initramfs 306,992 squashfs 339,968 cramfs 358,608 JFFS2 686,400 YAFFS2 577,537 romfs 701,440 ext2

60 主流嵌入式 Linux 文件系统 二级储存器 适合闪存使用的文件系统 针对 MTD 的文件系统 自带 FTL 的文件系统 其他嵌入式文件系统

61 Virtual File Systems Linux (like Unix) is designed to use any number of arbitrary file systems Provides uniform interface to file systems through the VFS (Virtual File System) Provides shared routines (like caching) Physical storage not necessary File systems implemented as kernel modules Most of the time (for embedded systems) compiled directly into the kernel Can be loaded as modules after kernel starts User space file systems: FUSE Fully functional file systems that run in user space Intriguing solution for embedded systems, more stable kernel

62 Virtual File Systems (Cont.) Virtual File Systems (VFS) on Linux provide an objectoriented way of implementing file systems VFS allows the same system call interface (the API) to be used for different types of file systems Separates file-system generic operations from implementation details Implementation can be one of many file systems types, or network file system Implements vnodes which hold inodes or network file details Then dispatches operation to appropriate file system implementation routines

63 Virtual File Systems (Cont.) The API is to the VFS interface, rather than any specific type of file system

64 Virtual File Systems (Cont.) For example, Linux has four object types: inode, file, superblock, dentry VFS defines set of operations on the objects that must be implemented Every object has a pointer to a function table Function table has addresses of routines to implement that function on that object For example: int open(...) Open a file int close(...) Close an already-open file ssize t read(...) Read from a file ssize t write(...) Write to a file int mmap(...) Memory-map a file

65 FUSE Part of the kernel starting at Kernel module User land helper programs and library Patches for Sample File Systems Media: DVD, Playlists, MythTV Dynamic Devices: USB Interesting: Database, Encrypted, GMail Language Bindings C, C++, Java, C#, Python, Perl, Ruby and some more More Info

66 FUSE, yes or no? Advantage: isolated from the kernel Disadvantage: high cost of space, time, etc. Not very space efficient or high-performance in its current release, so not super-useful for embedded applications. But keep your eyes peeled!

67 NFS Rarely used in production systems Great way for testing your board Relying on network in production is risky Not very fast on Linux, slow when using Cygwin as a server While a file system from a technical perspective, it is a protocol Makes some file system remotely accessible Negotiates privileges, what clients can access in the resource Can export any file system type for access over NFS (well, almost any)

68 内容提要 文件系统概念回顾 主流嵌入式 Linux 文件系统 如何选择合适的文件系统

69 如何选择合适的文件系统 针对合适的储存设备 针对合适的应用场景 参考选择

70 按适用储存设备分 On a block device the ext* family (such as ext2, ext3 and ext4), XFS, JFS, ReiserFS and btrfs. For raw flash without a flash translation layer (FTL) or Memory Technology Device (MTD) UBIFS, JFFS2, and YAFFS, among others. SquashFS is a common compressed read-only file system.

71 按应用场景 No storage initramfs, cramfs Ultra small storage squashfs, romfs Normal storage JFFS2, YAFFS2 Large storage UBIFS, LogFS

72 Recommended FS Read-Only squashfs Best compression, ability to control endianness and compression Flash JFFS2 The standard, compresses well, well-supported and rock solid, recent improvements in performance, too! Development NFS Small impact on kernel size, can configure as readonly so it looks like system

73 Reliability Consider flash memory erase cycles 10 3 (for MLC NAND) (NOR, SLC NAND) erase cycles Never mount a block FS on top of MTD/UBI device Otherwise, Consider bad block recovery Other same concerns on block file systems

74 主要资料来源 Silberschatz, Galvin and Gagne, Operating System Concepts, 9 th ed., Timesys, Survey of Linux Filesystems, Online Webinar, Wikipedia, Flash File Systems, webpages.