Lecture 023

File Operations

File: abstraction of persistent storage with logical grouping of data

Operation on Files:

• create

• read, write

• seek

• delete

• truncate

• append

• rename

• open: do expensive directory lookup, store partition, file number, allowed operations, cursor position (shared across processes)

File Structure

Program loader needs to know executables with magic number in first two bytes:

• ELF: 0x7F, 'E', 'L', 'F'

• '#!' for scripts

• if not, then it's a data file (array of bytes)

you can use file command to check the type of a data file

Access Methods

Sequential Access: read(), write() rewind() seek() Indexed Access: (for files contains records), index key to value

Directory Operations

Operations:

• lookup

• create

• delete

• rename

• iterate over

• scan (find)

• change notification (watch)

Hard link: instead of a tree, it's a DAG where we allow a file to own by multiple directories.

• any change to file reflects in all directories

• delete a file only removes the link, the file is deleted when all links are removed

• hard link only apply to files, not directories (to avoid cycles)

Soft link: symbolic link, short cut

• if somebody delete the original file, the soft link becomes dangling

• can link to directories, but not across file systems

Mounting: attach a file system to a directory, the directory becomes the root of the file system

User/Group: uid/gid, used for access control

• access mode: read, write, execute (append, delete, list, lock)

• access control list (ACL): (user, mode) tuples, cumbersome to store/view/manage

• capability: given user a list of (file, access keys) tuple, hard to revoke

• typical: every file has permission for owner, group, and other (and rwx for each), we have 16 bits to store permissions, user#, group# (which is tight, will not allow many users/groups)