Smartphone Operating System (779883), страница 40
Текст из файла (страница 40)
We would look up the namereadme.txt (see Figure 8.5) and get the block number of the first fileblock. Assuming a standard cluster size of 512 bytes, we would needtwo blocks. We would read the first block from the disk and display itsIMPLEMENTATION OF A FILE SYSTEM185Directory entryReadme.txtName...1045First block2067321End of fileFile allocationtableFigure 8.5 Files stored with the file-allocation tablecontents. We then need the table entry for the first block to get the secondblock. We would read the second block and display the remaining bytes.Many storage media support some version of the FAT file system.FAT12 is used for floppy disks; FAT16 is used on most removable media(USB flash drives, for example).
FAT32 is compatible with storage mediaused by Microsoft Windows 2000 and XP.NTFSThe New Technology File System – otherwise known as NTFS – debutedwith Microsoft Windows NT. It supports many innovations over the FAT16system used by previous Microsoft Windows versions: compression, filelevel security, larger partitions and RAID.
In addition, NTFS supportsencryption of file-system data. One of the main features of NTFS isexceptional fault tolerance, because it is a transactional file system.NTFS does away with the file-allocation table and completely changesthe way partitions are formatted. A master file table (analogous to the FAT)186FILE SYSTEMS AND STORAGEis stored on disk and an address to it is stored in the boot sector. The bootsector contains code that starts the boot-up process. Most interestingly,the boot sector has as its first bytes a jump instruction that enables a jumpto where the bootstrap code is located in the boot sector. This means thatdata can be large or small and the operating system can always find bootcode.NTFS is the favored file system on Microsoft Windows installations.Unix File Systems: VFS and UFSUnix generally uses two different kinds of file system.
A virtual file system(VFS) is an abstract file system, with abstract interfaces to commonly usedsystem calls. Programmers and applications make VFS calls to access filesystem facilities. The VFS calls are then implemented by real file-systemimplementations. Two examples of real Unix file systems are the Unix Filesystem (UFS) and the Network File system (NFS). VFS is effective becauseUnix is designed to use many different file-system implementations.Implementations of abstract VFS calls are loaded dynamically, and callsare made using a specific type of implemented file system.The UFS, also known as the Berkeley Fast File system, is used by manyUnix implementations.
Each partition starts with space reserved for bootblocks, addresses that point to operating system code that resides onthe storage medium. The next space is called a superblock ; it containsnumbers that identify the partition and parameters that can be altered totune the behavior of the partition. The remainder of the storage spaceis organized into groups of fixed size. Each group contains a duplicatecopy of the superblock; a group header, containing statistics, pointersto free blocks and tunable parameters; a number of inodes that containinformation about files; and blocks of data that contain file content.Inodes are the heart of a UFS implementation.
Inodes are a combinationof file information – name, ownership, times of access, etc. – and pointersto blocks of file content. In fact, inodes are directly analogous to FCBs.Inodes implement file data pointers in an interesting way. They store 12direct blocks, which point directly to blocks that hold file content. The13th entry is an indirect block, which points to a file block on the storagemedium that holds the addresses of other file blocks.
The 14th entry is adouble indirect block, which points to a file block whose addresses pointto indirect blocks. The last entry is a triple indirect block, which has threelevels of indirection built into it.The indirection built into the design of an inode needs a bit of explaining. If we assume each file block holds 15 pointers, then each inode canIMPLEMENTATION OF A FILE SYSTEM187access 3 627 file blocks.1 Consider the space required if each inode held3 627 file block addresses: 14.5 KB of space to access each file. If thefile system were to be filled with predominantly small files, then manyof the addresses would go unused. In the inode scheme, with a standard4 096-byte block, files up to 48 KB in size can be accessed directly fromthe inode.
Files that are between 48 KB and 108 KB take only one moreblock in the inode, but require an extra file access to read the file blockwith the actual disk addresses. This scheme favors smaller files, allowingaccess to large files – up to 4 GB – to have a slight penalty.Remote File SystemsServers usually have file systems that they set up for other computers touse over a network. This type of file service allows client computers touse the file systems of servers as if they were local.A server file system is (obviously) local to the server it is hostedon.
It is exported via network protocols to other computers. Two ofthe most widely used protocols are the Network File Service (NFS) andServer Message Block (SMB) file systems. The former originated on SunMicrosystems SunOS operating system and has been implemented formost other operating systems. The latter originated with Microsoft forits operating systems and has also been implemented for many otheroperating systems.The goal of using remote file systems is to allow the user or applicationnot to see the difference between a locally implemented file system and aremote file system.
This is where the abstraction of using file APIs is mostimportant. Consider VFS from Unix. Under VFS, the same API is used nomatter if the underlying file system is UFS or NFS. The open() functionis called no matter what. The underlying file system provides a specificimplementation of the open() function.Other Interesting File System ImplementationsThere are many file systems that have been developed over the yearsthat operating systems have used.
There have been several recent innovations that are in use. Log-structured file systems write every file-system1This is computed as follows: 12 direct blocks; one indirect block, which adds 15 more;one doubly indirect block, which adds 15 indirect blocks, each of which adds 15 more;and finally one triply indirect block, adding 15 x15 x15 blocks. The result is 12 + 15 + 225+ 3 375 = 3 627.188FILE SYSTEMS AND STORAGEmodification to a file, allowing them to be replayed and analyzed.
Avariant of this type of file system – the journaling file system – is in usein Linux devices (the ext3 file system). Other file systems in use todayare the Universal Disk Format system for DVDs and CD-R/RWs and theHierarchical File System used by MacOS.8.3 File Systems on Mobile PhonesIn terms of file systems, mobile phone operating systems have many ofthe requirements of desktop operating systems. Most are implemented in32-bit environments; most allow users to give arbitrary names to files;most store many files that require some kind of organized structure. Thismeans that a hierarchical directory-based file system is desirable.
Andwhile designers of mobile operating systems have many choices for filesystems, one more characteristic influences their choice: most mobilephones have storage media that can be shared with a Microsoft Windowsenvironment.If mobile phone systems did not have removable media, then anyfile system would be usable. In systems that use flash memory, there arespecial circumstances to consider. Block sizes are typically from 512 bytesto 2 048 bytes.
Flash memory cannot simply overwrite memory; it musterase first, then write. In addition, the unit of erasure is rather coarse:individual bytes cannot be erased; entire blocks must be erased at a time.Erase times for flash memory is relatively long.To accommodate these characteristics, flash memory works best whenthere are specifically designed file systems that spread writes over themedia and deal with the long erase times. The basic concept is that whenthe flash store is to be updated, the file system writes a new copy of thechanged data over to a fresh block, remaps the file pointers, then erasesthe old block later when it has time.One of the earliest flash file systems was Microsoft’s FFS2 for usewith MS-DOS in the early 1990s.
When the PCMCIA industry groupapproved the Flash Translation Layer specification for flash memory in1994, flash devices could look like a FAT file system. Linux also hasspecially designed file systems, from the Journaling Flash File System( JFFS) to the Yet Another Flash Filing System (YAFFS).However, mobile phone platforms must share their media with othercomputers, which demands that some form of compatibility be in place.SECURITY189Most often, FAT file systems are used. Specifically, FAT16 is used for itsshorter allocation table (than FAT32) and for its reduced need for longfiles.Being a mobile smartphone operating system, Symbian OS needs toimplement at least the FAT16 file system.
Indeed, it provides support forFAT16 and uses that file system for most of its storage media. However,the Symbian OS file-server implementation is built on an abstractionmuch like Unix’s VFS. Object orientation allows objects that implementvarious operating systems to be plugged into the Symbian OS file server,thus allowing many different file-system implementations to be used.Different implementations may even co-exist in the same file server.Implementations of NFS and SMB file systems have been created forSymbian OS.8.4 SecuritySecurity is very important for file systems.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
