User identifier

Unix-like operating systems identify a user within the kernel by a value called a user identifier, often abbreviated to user ID or UID. The UID, along with the group identifier (GID) and other access control criteria, is used to determine which system resources a user can access. The password file maps textual user names to UIDs, but in the kernel, only UIDs are used. UIDs are stored in the inodes of the Unix file system, running processes, tar archives, and the now-obsolete Network Information Service. In POSIX-compliant environments, the command-line command id gives the current user's UID, as well as more information such as the user name, primary user group and group identifier (GID).

Process attributes

The POSIX standard introduced three different UID fields into the process descriptor table, to allow privileged processes to take on different roles dynamically:

Effective user ID

The effective UID (euid) of a process is used for most access checks. It is also used as the owner for files created by that process. The effective GID (egid) of a process also affects access control and may also affect file creation, depending on the semantics of the specific kernel implementation in use and possibly the mount options used. According to BSD Unix semantics, the group ownership given a newly created file is unconditionally inherited from the group ownership of the directory in which it is created. According to AT&T UNIX System V semantics (also adopted by Linux variants) newly created files will normally be given the group ownership of the egid of the process that creates them. Most filesystems implement a method to select whether BSD or AT&T semantics should be used regarding group ownership of newly created files, BSD semantics is selected for specific directories in case that the S_ISGID (s-gid) permission is set.^[1]

File system user ID

Linux also has a file system user ID (fsuid) which is used explicitly for access control to the file system. It matches the euid unless explicitly set otherwise. It may be root's user ID only if ruid, suid, or euid is root. Whenever the euid is changed, the change is propagated to the fsuid.

The intent of fsuid is to permit programs (e.g., the NFS server) to limit themselves to the file system rights of some given uid without giving that uid permission to send them signals. Since kernel 2.0, the existence of fsuid is no longer necessary because Linux adheres to SUSv3 rules for sending signals, but fsuid remains for compatibility reasons.^[2]

Saved user ID

The saved user ID (suid) is used when a program running with elevated privileges needs to temporarily do some unprivileged work: it changes its effective user ID from a privileged value (typically root) to some unprivileged one, and this triggers a copy of the privileged user ID to the saved user ID slot.^[3] Later, it can set its effective user ID back to the saved user ID (an unprivileged process can only set its effective user ID to three values: its real user ID, its saved user ID, and its effective user ID—i.e., unchanged) to resume its privileges.

Real user ID

The real UID (ruid) and real GID (rgid) identify the real owner of the process and affect the permissions for sending signals. A process without superuser privilege can signal another process only if the sender’s real or effective UID matches the real or saved UID of the receiver.^[2] Since child processes inherit the credentials from the parent, they can signal each other.

Conventions

Type

POSIX requires the UID to be an integer type. Most Unix-like operating systems represent the UID as an unsigned integer. The size of UID values varies amongst different systems; some UNIX OS's used 15-bit values, allowing values up to 32767, while others such as Linux (before version 2.4) supported 16-bit UIDs, making 65536 unique IDs possible. The majority of modern Unix-like systems (e.g., Solaris-2.0 in 1990, Linux 2.4 in 2001) have switched to 32-bit UIDs, allowing 4,294,967,296 (2³²) unique IDs.

Reserved ranges

The Linux Standard Base Core Specification specifies that UID values in the range 0 to 99 should be statically allocated by the system, and shall not be created by applications, while UIDs from 100 to 499 should be reserved for dynamic allocation by system administrators and post install scripts.^[4]

On FreeBSD, porters who need a UID for their package can pick a free one from the range 50 to 999 and then register this static allocation in ports/UIDs.^[5]

Some POSIX systems allocate UIDs for new users starting from 500 (OS X, Red Hat Enterprise Linux), others start at 1000 (openSUSE, Debian^[6]). On many Linux systems, these ranges are specified in /etc/login.defs, for useradd and similar tools.

Central UID allocations in enterprise networks (e.g., via LDAP and NFS servers) may limit themselves to using only UID numbers well above 1000, to avoid potential conflicts with UIDs locally allocated on client computers. NFSv4 can help avoid numeric identifier collisions, by identifying users (and groups) in protocol packets using "user@domain" names rather than integer numbers, at the expense of additional translation steps.

Special values

• 0: The superuser normally has a UID of zero (0).^[7]
• −1: The value (uid_t) -1 is reserved by POSIX to identify an omitted argument.^[8]
• Nobody: Historically, the user “nobody” was assigned UID -2 by several operating systems, although other values such as 2¹⁵−1 = 32,767 are also in use, such as by OpenBSD.^[9] For compatibility between 16-bit and 32-bit UIDs, many Linux distributions now set it to be 2¹⁶−2 = 65,534; the Linux kernel defaults to returning this value when a 32-bit UID does not fit into the return value of the 16-bit system calls.^[10] An alternative convention assigns the last UID of the range statically allocated for system use (0-99) to nobody: 99.

See also

• setuid
• Sticky bit
• Group identifier
• Process identifier
• File system permissions
• Open (system call)
• Mount (Unix)
• FAT access rights
• Security Identifier (SID) – the Windows NT equivalent

References