File Systems: Interfaces and Operations - Prof. Michael W. Hicks, Study notes of Operating Systems

Download File Systems: Interfaces and Operations - Prof. Michael W. Hicks and more Study notes Operating Systems in PDF only on Docsity!

CMSC 412

Filesystems: Interfaces

Announcements

• Reading
  • Today : Chapter 10
  • Next time : Chapter 11

File Abstraction

• What is a file?
  • A named collection of information stored on secondary storage
• Properties of a file
  • non-volatile (persistent)
  • can read, store, or update it
  • has meta-data to describe attributes of the file
  • May be structured or unstructured

File Attributes

• name : a way to describe the file
• type : some information about what is stored in the file
• location : how to find the file on disk
• size : number of bytes
• protection : access control
  • may be different for read, write, execute, append, etc.
• time : access, modification, creation
• version : how many times the file changed

Write

• Indicate what file to write (either name of handle)
• Provide data to write
• Specify where to write the data within the file - generally this is implicit (file pointer) - Advances file pointer afterward - could be explicit ( direct access ) - Specify relative block number - Or use some kind of index file

Read

• Indicate what file to read (either name or handle)
• Provide place to put information read
• Indicate how much to read
• Specify where to write the data within the file - Implicit or explicit, as with writing

Fsync, Close

• fsync:
  • synchronize disk version with in-core version
  • ensures any previous writes to the file are stored on disk
• close
  • unlock the file (if locked when opened)
  • update meta data about time
  • free system resources (file descriptors, buffers)

Seek, Delete, Truncate

• seek
  • move the implicit file pointer to a new offset in the file
• delete
  • remove named file
• truncate
  • remove the data in the file from the current position to end

Example of File Types

• Macintosh
  • has a file type that is part of file meta-data
  • also has an application associated with each file type
• Windows 95/NT
  • has a file type in the extension of the file name
  • has a table (per user) to map extensions to applications
• Unix
  • can use last part of filename like an extension
  • applications can decide what (if anything) to do with it

Filesystems

• Provides a namespace for files via directories - Each file system part of a disk partition
• Can store files of variable size
• Provides protection by restricting access to files based on permissions

Directory

• Collection of files
• Operations
  • Search for a file
  • Create a file
  • Delete a file
  • List a directory
  • Rename a file
  • Traverse the file system

Single Directory Structure

• All files are in a single global namespace
• Simple, but having all of the files in one name space is awkward - lots of files to sort through - different users would have to coordinate file names - each file has to have a unique name

Tree Directories

a b mwh user1 user cc (^) vi gdb one^ to x y system users mail z

Acylic Graph Directories

• Users can share directories and files a b mwh user1 user cc vi two one^ to x y system users mail z

Acyclic Graph Issues

• Same file may have several names
  • absolute path name differs, but file is the same
  • similar to memory aliases in prog. languages
• Deletion f one user deletes a shared file
  • It is deleted for all users; dangling links
  • It stays until the last referent deletes it
    • Maintains a reference count to determine this
• Programs to walk the DAG need to be aware
  • disk usage utilities, backup utilities
• What if the graph was not acyclic?
  • Would need garbage collection for dead files

Symbolic vs. Hard links

• Hard link
  • Metadata in the file system “points” to the file or directory entry directly - Dangling links not permitted
• Symbolic link
  • Use a level of indirection: encoded as a path in the file system - Dangling links permitted: if the file is removed, then opening the link will fail

Mount Point

File Protection

• File sharing implies need of protection :
  • How to give access to some users and not others?
• Access types:
  • read, write, execute, append, delete, list
  • rename: often based on protection of directory
  • copy: usually the same as read

Access Policies

• Access lists
  • list for each user for each file the permitted operations
• Groups
  • enumerate users in a list called a group
  • same protection to all members of the group
  • depending on system:
    • files may be in one or many groups
    • users may be in one or many groups
• Per-file passwords
  • tedious and a security problem

UNIX File Protection

• Each file has three classifications
  • user: the user who owns the file
  • group: a named group of other users
  • world: all others
• Each file has three access types:
  • read, write, execute
• Three additional bits
  • Sticky bit
    • leave executable in memory after is done
  • Setuid, Setgid
    • run the program with the uid/gid of the file’s owner
    • used to provide extra privilege to some processes
      • example: passwd command

AFS File Protection

• Each Directory has an ACL (but no file ACL)
  • applies to all files in a directory, and subdirectories (but can override)
  • 7 file access types:
    • read, write, lookup, delete, insert, lock (k), administer
  • ACLs apply to a user or a group
  • ACL may contain negative rights
    • “everyone but Joe Smith may read this file”
• Groups
  • each user can create a fixed number of groups
    • users can manage their own groups
• Cells: collections of computers
  • csic, wam, … creates networked namespace

Remote File Systems

• Uses networking to allow file system access between systems - Manually via programs like FTP - Automatically using distributed file systems - Semi-automatically via the world wide web
• Client-server model allows clients to mount remote file systems from servers - Server can serve multiple clients - Client and user-on-client identification is insecure or complicated - NFS is standard UNIX file sharing protocol - CIFS is standard Windows protocol - Standard operating system file calls are translated into remote calls

Effect of Updates to Shared Files

• UNIX

• writes are visible immediately
• have a mode to permit processes to share file pointers
• AFS
• open/close semantics
• “copy” the file on open
• write-back on close
• Immutable files
• once “published”, the file never changes
• usually done by attaching a version # to the filename
• new versions of the file are given a new name

Raw Disks vs. Filesystems

• Can implement file systems with raw disks , which can be viewed as: - a linear array of fixed sized units of allocation, called blocks - assume that blocks are error free (for now) - typical block size is 512 to 4096 bytes - can update a block in place, but must write the entire block - can access any block in any desired order - blocks must be read as a unit - for performance reasons may care about “near” vs. “far” blocks (but that is covered in a future lecture)