File Systems cs550 Operating Systems David Monismith

Files File - sequence of bytes File attributes Size Location Type Permissions Time stamps Folders Path name specifies sequence of folders users must traverse to travel down the tree to the file

E.g. path for B2 is /abc/def/B2 absolute path for B2 abc / \ | | def A1 | | B1 B2 Files also have relative paths – Sequence of folders users must traverse starting at an intermediate place Files often have symbolic links or shortcuts, too. These take you directly to another folder where the file resides – Fast way to get to the file you want Files

File Access Methods 1.Sequential 2.Random What types of storage devices allow for these types of access? Operations including open, close, read, and write allow interaction with files Open

File Access - Open Needed to read or write OS needs to know where file is located Requires directory traversal Don't want to do this with every read and write, because it requires disk accesses

Access Methods (Contd..) Require open call to avoid overhead Do directory traversal once Cache directory entry in main memory – Get fast access to file – Call to open returns a "file handle" identifying the cached directory entry Typically must identify whether you are reading or writing and whether sequential or random access will be used Writing can start at the end of the file (append) or the beginning

File Access Methods - Close Several functions exist to perform a close operation The operation destroys a cached directory entry Ensure all data has been written to the file (it may be buffered) If on removable media, makes sure media is ready to be removed

File Access - Reading Read data from file into memory space of process requesting data File Mgmt System needs several pieces of information Which file? Use handle from open call Where is the data in the file? Typically don't read a whole file at once Instead, read data blocks Indicate what portion of the file to read Where will data be stored? Typically in a buffer in memory. How much data to read Typically a system level read call is as follows: read(handle, file position, buffer, length)

File Access - Write Similar to read (but opposite) Write data from process memory space to file Again File Mgmt. System needs several pieces of information Which file? Need file handle Where to store data? Need to write portions of the file (blocks) Where is the buffer that stores the data in memory? How much data? Typically done with following pseudo code write(handle, file position, buffer, length)

Sequential Access If file is open for sequential access, file mgmt subsystem keeps track of current file position for reading and writing System maintains a file position pointer for position of next read or write Often set as follows:  Set to zero to start reading or writing at the beginning of a file  Set to the end of the file to append data Pointer is incremented by the amount of data read or written after each read or write

Streams, Pipes, and Redirection We looked at (used) many of these but we haven't analyzed them Stream - flow of data bytes (sequentially) into the process (read) and out of the process (write). Used by java, c++, and windows.net

Pipes A pipe is a dynamic connection established between two processes A pipe has a process at one end that writes and process at the other end reads Allows for interprocess communication Data flows one way Pipes can improve system performance by not needing to use a file And by allowing for buffered data to be sent from one process to another as the initial process is still reading it Pipes often use a buffer in main memory as a temporary holding place for data (actually an implementation of producer consumer problem)

When a process executes the file management system creates 3 standard I/O files  stdin - takes input from the user keyboard (typically)  stdout - typically writes to the screen  stderr - used for process error messages Each file is a pipe that is connected to a system module that communicates with the appropriate device The command line interface supports I/O redirection Using redirection, the user can request the CLI  Take data from a file or another process using redirection instead of stdin  Send data to another file or process instead of stdout or stderr Standard I/O Files

File redirection means sending input data to or forwarding output data to another file as input or output, respectively, via the command line. An example of this follows below dir > tmp.txt more < tmp.txt dir | more Redirection

Other I/O Calls Many other system I/O calls exist including flush directory functions create file delete file rename file file exists directory list get and set attributes

File space allocation Allocate space for files Simple to use contiguous allocation Optimal performance for application reading/writing the file Problems Need to know how much space the file will take up After long run time file system becomes cluttered Need to move files around Cluster allocation Used to solve problem Allocate chunks of data and link them together Can also have indexed clusters

Calculating read and write addresses Typically use file pointer Use offset to get to address More complicated when using clusters Data requests may cross cluster boundaries Free space management Keep track of clusters allocated to each file Use linked list Or use bit mapping Each bit in the bitmap represents a cluster  On – used  Off - free File space allocation

Disk fragmentation External File broken into many clusters scattered across disk More clusters = poor performance Can allocate larger clusters, but need more defragmentation i.e. cluster rearrangement Internal If using large clusters, small files use lots of disk space Waste may be a concern Next time: Real world file systems