29. Users and Groups

Every user who can log in on the system is identified by a unique number called the user ID. Each process has an effective user ID which says which user's access permissions it has.

Users are classified into groups for access control purposes. Each process has one or more group ID values which say which groups the process can use for access to files.

The effective user and group IDs of a process collectively form its persona. This determines which files the process can access. Normally, a process inherits its persona from the parent process, but under special circumstances a process can change its persona and thus change its access permissions.

Each file in the system also has a user ID and a group ID. Access control works by comparing the user and group IDs of the file with those of the running process.

The system keeps a database of all the registered users, and another database of all the defined groups. There are library functions you can use to examine these databases.

29.1 User and Group IDs

Each user account on a computer system is identified by a user name (or login name) and user ID. Normally, each user name has a unique user ID, but it is possible for several login names to have the same user ID. The user names and corresponding user IDs are stored in a data base which you can access as described in User Database.

Users are classified in groups. Each user name belongs to one default group and may also belong to any number of supplementary groups. Users who are members of the same group can share resources (such as files) that are not accessible to users who are not a member of that group. Each group has a group name and group ID. See section Group Database, for how to find information about a group ID or group name.

29.2 The Persona of a Process

At any time, each process has an effective user ID, a effective group ID, and a set of supplementary group IDs. These IDs determine the privileges of the process. They are collectively called the persona of the process, because they determine “who it is” for purposes of access control.

Your login shell starts out with a persona which consists of your user ID, your default group ID, and your supplementary group IDs (if you are in more than one group). In normal circumstances, all your other processes inherit these values.

A process also has a real user ID which identifies the user who created the process, and a real group ID which identifies that user's default group. These values do not play a role in access control, so we do not consider them part of the persona. But they are also important.

Both the real and effective user ID can be changed during the lifetime of a process. See section Why Change the Persona of a Process?.

For details on how a process's effective user ID and group IDs affect its permission to access files, see How Your Access to a File is Decided.

The effective user ID of a process also controls permissions for sending signals using the kill function. See section Signaling Another Process.

Finally, there are many operations which can only be performed by a process whose effective user ID is zero. A process with this user ID is a privileged process. Commonly the user name root is associated with user ID 0, but there may be other user names with this ID.

29.3 Why Change the Persona of a Process?

The most obvious situation where it is necessary for a process to change its user and/or group IDs is the login program. When login starts running, its user ID is root. Its job is to start a shell whose user and group IDs are those of the user who is logging in. (To accomplish this fully, login must set the real user and group IDs as well as its persona. But this is a special case.)

The more common case of changing persona is when an ordinary user program needs access to a resource that wouldn't ordinarily be accessible to the user actually running it.

For example, you may have a file that is controlled by your program but that shouldn't be read or modified directly by other users, either because it implements some kind of locking protocol, or because you want to preserve the integrity or privacy of the information it contains. This kind of restricted access can be implemented by having the program change its effective user or group ID to match that of the resource.

Thus, imagine a game program that saves scores in a file. The game program itself needs to be able to update this file no matter who is running it, but if users can write the file without going through the game, they can give themselves any scores they like. Some people consider this undesirable, or even reprehensible. It can be prevented by creating a new user ID and login name (say, games) to own the scores file, and make the file writable only by this user. Then, when the game program wants to update this file, it can change its effective user ID to be that for games. In effect, the program must adopt the persona of games so it can write the scores file.

29.4 How an Application Can Change Persona

The ability to change the persona of a process can be a source of unintentional privacy violations, or even intentional abuse. Because of the potential for problems, changing persona is restricted to special circumstances.

You can't arbitrarily set your user ID or group ID to anything you want; only privileged processes can do that. Instead, the normal way for a program to change its persona is that it has been set up in advance to change to a particular user or group. This is the function of the setuid and setgid bits of a file's access mode. See section The Mode Bits for Access Permission.

When the setuid bit of an executable file is on, executing that file gives the process a third user ID: the file user ID. This ID is set to the owner ID of the file. The system then changes the effective user ID to the file user ID. The real user ID remains as it was. Likewise, if the setgid bit is on, the process is given a file group ID equal to the group ID of the file, and its effective group ID is changed to the file group ID.

If a process has a file ID (user or group), then it can at any time change its effective ID to its real ID and back to its file ID. Programs use this feature to relinquish their special privileges except when they actually need them. This makes it less likely that they can be tricked into doing something inappropriate with their privileges.

Portability Note: Older systems do not have file IDs. To determine if a system has this feature, you can test the compiler define _POSIX_SAVED_IDS. (In the POSIX standard, file IDs are known as saved IDs.)

See section File Attributes, for a more general discussion of file modes and accessibility.

29.5 Reading the Persona of a Process

Here are detailed descriptions of the functions for reading the user and group IDs of a process, both real and effective. To use these facilities, you must include the header files ‘sys/types.h’ and ‘unistd.h’.

Data Type: uid_t

This is an integer data type used to represent user IDs. In the GNU library, this is an alias for unsigned int.

Data Type: gid_t

This is an integer data type used to represent group IDs. In the GNU library, this is an alias for unsigned int.

Function: uid_t getuid (void)

The getuid function returns the real user ID of the process.

Function: gid_t getgid (void)

The getgid function returns the real group ID of the process.

Function: uid_t geteuid (void)

The geteuid function returns the effective user ID of the process.

Function: gid_t getegid (void)

The getegid function returns the effective group ID of the process.

Function: int getgroups (int count, gid_t *groups)

The getgroups function is used to inquire about the supplementary group IDs of the process. Up to count of these group IDs are stored in the array groups; the return value from the function is the number of group IDs actually stored. If count is smaller than the total number of supplementary group IDs, then getgroups returns a value of -1 and errno is set to EINVAL.

If count is zero, then getgroups just returns the total number of supplementary group IDs. On systems that do not support supplementary groups, this will always be zero.

Here's how to use getgroups to read all the supplementary group IDs:

gid_t * read_all_groups (void) { int ngroups = getgroups (0, NULL); gid_t *groups = (gid_t *) xmalloc (ngroups * sizeof (gid_t)); int val = getgroups (ngroups, groups); if (val < 0) { free (groups); return NULL; } return groups; }

29.6 Setting the User ID

This section describes the functions for altering the user ID (real and/or effective) of a process. To use these facilities, you must include the header files ‘sys/types.h’ and ‘unistd.h’.

Function: int seteuid (uid_t neweuid)

This function sets the effective user ID of a process to newuid, provided that the process is allowed to change its effective user ID. A privileged process (effective user ID zero) can change its effective user ID to any legal value. An unprivileged process with a file user ID can change its effective user ID to its real user ID or to its file user ID. Otherwise, a process may not change its effective user ID at all.

The seteuid function returns a value of 0 to indicate successful completion, and a value of -1 to indicate an error. The following errno error conditions are defined for this function:

EINVAL

The value of the newuid argument is invalid.

EPERM

The process may not change to the specified ID.

Older systems (those without the _POSIX_SAVED_IDS feature) do not have this function.

Function: int setuid (uid_t newuid)

If the calling process is privileged, this function sets both the real and effective user ID of the process to newuid. It also deletes the file user ID of the process, if any. newuid may be any legal value. (Once this has been done, there is no way to recover the old effective user ID.)

If the process is not privileged, and the system supports the _POSIX_SAVED_IDS feature, then this function behaves like seteuid.

The return values and error conditions are the same as for seteuid.

Function: int setreuid (uid_t ruid, uid_t euid)

This function sets the real user ID of the process to ruid and the effective user ID to euid. If ruid is -1, it means not to change the real user ID; likewise if euid is -1, it means not to change the effective user ID.

The setreuid function exists for compatibility with 4.3 BSD Unix, which does not support file IDs. You can use this function to swap the effective and real user IDs of the process. (Privileged processes are not limited to this particular usage.) If file IDs are supported, you should use that feature instead of this function. See section Enabling and Disabling Setuid Access.

The return value is 0 on success and -1 on failure. The following errno error conditions are defined for this function:

EPERM

The process does not have the appropriate privileges; you do not have permission to change to the specified ID.

29.7 Setting the Group IDs

This section describes the functions for altering the group IDs (real and effective) of a process. To use these facilities, you must include the header files ‘sys/types.h’ and ‘unistd.h’.

Function: int setegid (gid_t newgid)

This function sets the effective group ID of the process to newgid, provided that the process is allowed to change its group ID. Just as with seteuid, if the process is privileged it may change its effective group ID to any value; if it isn't, but it has a file group ID, then it may change to its real group ID or file group ID; otherwise it may not change its effective group ID.

Note that a process is only privileged if its effective user ID is zero. The effective group ID only affects access permissions.

The return values and error conditions for setegid are the same as those for seteuid.

This function is only present if _POSIX_SAVED_IDS is defined.

Function: int setgid (gid_t newgid)

This function sets both the real and effective group ID of the process to newgid, provided that the process is privileged. It also deletes the file group ID, if any.

If the process is not privileged, then setgid behaves like setegid.

The return values and error conditions for setgid are the same as those for seteuid.

Function: int setregid (gid_t rgid, gid_t egid)

This function sets the real group ID of the process to rgid and the effective group ID to egid. If rgid is -1, it means not to change the real group ID; likewise if egid is -1, it means not to change the effective group ID.

The setregid function is provided for compatibility with 4.3 BSD Unix, which does not support file IDs. You can use this function to swap the effective and real group IDs of the process. (Privileged processes are not limited to this usage.) If file IDs are supported, you should use that feature instead of using this function. See section Enabling and Disabling Setuid Access.

The return values and error conditions for setregid are the same as those for setreuid.

setuid and setgid behave differently depending on whether the effective user ID at the time is zero. If it is not zero, they behave like seteuid and setegid. If it is, they change both effective and real IDs and delete the file ID. To avoid confusion, we recommend you always use seteuid and setegid except when you know the effective user ID is zero and your intent is to change the persona permanently. This case is rare—most of the programs that need it, such as login and su, have already been written.

Note that if your program is setuid to some user other than root, there is no way to drop privileges permanently.

The system also lets privileged processes change their supplementary group IDs. To use setgroups or initgroups, your programs should include the header file ‘grp.h’.

Function: int setgroups (size_t count, gid_t *groups)

This function sets the process's supplementary group IDs. It can only be called from privileged processes. The count argument specifies the number of group IDs in the array groups.

This function returns 0 if successful and -1 on error. The following errno error conditions are defined for this function:

EPERM

The calling process is not privileged.

Function: int initgroups (const char *user, gid_t group)

The initgroups function sets the process's supplementary group IDs to be the normal default for the user name user. The group group is automatically included.

This function works by scanning the group database for all the groups user belongs to. It then calls setgroups with the list it has constructed.

The return values and error conditions are the same as for setgroups.

If you are interested in the groups a particular user belongs to, but do not want to change the process's supplementary group IDs, you can use getgrouplist. To use getgrouplist, your programs should include the header file ‘grp.h’.

Function: int getgrouplist (const char *user, gid_t group, gid_t *groups, int *ngroups)

The getgrouplist function scans the group database for all the groups user belongs to. Up to *ngroups group IDs corresponding to these groups are stored in the array groups; the return value from the function is the number of group IDs actually stored. If *ngroups is smaller than the total number of groups found, then getgrouplist returns a value of -1 and stores the actual number of groups in *ngroups. The group group is automatically included in the list of groups returned by getgrouplist.

Here's how to use getgrouplist to read all supplementary groups for user:

gid_t * supplementary_groups (char *user) { int ngroups = 16; gid_t *groups = (gid_t *) xmalloc (ngroups * sizeof (gid_t)); struct passwd *pw = getpwnam (user); if (pw == NULL) return NULL; if (getgrouplist (pw->pw_name, pw->pw_gid, groups, &ngroups) < 0) { groups = xrealloc (ngroups * sizeof (gid_t)); getgrouplist (pw->pw_name, pw->pw_gid, groups, &ngroups); } return groups; }

29.8 Enabling and Disabling Setuid Access

A typical setuid program does not need its special access all of the time. It's a good idea to turn off this access when it isn't needed, so it can't possibly give unintended access.

If the system supports the _POSIX_SAVED_IDS feature, you can accomplish this with seteuid. When the game program starts, its real user ID is jdoe, its effective user ID is games, and its saved user ID is also games. The program should record both user ID values once at the beginning, like this:

user_user_id = getuid ();
game_user_id = geteuid ();

Then it can turn off game file access with

seteuid (user_user_id);

and turn it on with

seteuid (game_user_id);

Throughout this process, the real user ID remains jdoe and the file user ID remains games, so the program can always set its effective user ID to either one.

On other systems that don't support file user IDs, you can turn setuid access on and off by using setreuid to swap the real and effective user IDs of the process, as follows:

setreuid (geteuid (), getuid ());

This special case is always allowed—it cannot fail.

Why does this have the effect of toggling the setuid access? Suppose a game program has just started, and its real user ID is jdoe while its effective user ID is games. In this state, the game can write the scores file. If it swaps the two uids, the real becomes games and the effective becomes jdoe; now the program has only jdoe access. Another swap brings games back to the effective user ID and restores access to the scores file.

In order to handle both kinds of systems, test for the saved user ID feature with a preprocessor conditional, like this:

#ifdef _POSIX_SAVED_IDS
  seteuid (user_user_id);
#else
  setreuid (geteuid (), getuid ());
#endif

29.9 Setuid Program Example

Here's an example showing how to set up a program that changes its effective user ID.

This is part of a game program called caber-toss that manipulates a file ‘scores’ that should be writable only by the game program itself. The program assumes that its executable file will be installed with the setuid bit set and owned by the same user as the ‘scores’ file. Typically, a system administrator will set up an account like games for this purpose.

The executable file is given mode 4755, so that doing an ‘ls -l’ on it produces output like:

-rwsr-xr-x   1 games    184422 Jul 30 15:17 caber-toss

The setuid bit shows up in the file modes as the ‘s’.

The scores file is given mode 644, and doing an ‘ls -l’ on it shows:

-rw-r--r--  1 games           0 Jul 31 15:33 scores

Here are the parts of the program that show how to set up the changed user ID. This program is conditionalized so that it makes use of the file IDs feature if it is supported, and otherwise uses setreuid to swap the effective and real user IDs.

#include <stdio.h>
#include <sys/types.h>
#include <unistd.h>
#include <stdlib.h>


/* Remember the effective and real UIDs. */

static uid_t euid, ruid;


/* Restore the effective UID to its original value. */

void
do_setuid (void)
{
  int status;

#ifdef _POSIX_SAVED_IDS
  status = seteuid (euid);
#else
  status = setreuid (ruid, euid);
#endif
  if (status < 0) {
    fprintf (stderr, "Couldn't set uid.\n");
    exit (status);
    }
}


/* Set the effective UID to the real UID. */

void
undo_setuid (void)
{
  int status;

#ifdef _POSIX_SAVED_IDS
  status = seteuid (ruid);
#else
  status = setreuid (euid, ruid);
#endif
  if (status < 0) {
    fprintf (stderr, "Couldn't set uid.\n");
    exit (status);
    }
}

/* Main program. */

int
main (void)
{
  /* Remember the real and effective user IDs.  */
  ruid = getuid ();
  euid = geteuid ();
  undo_setuid ();

  /* Do the game and record the score.  */
  …
}

Notice how the first thing the main function does is to set the effective user ID back to the real user ID. This is so that any other file accesses that are performed while the user is playing the game use the real user ID for determining permissions. Only when the program needs to open the scores file does it switch back to the file user ID, like this:

/* Record the score. */

int
record_score (int score)
{
  FILE *stream;
  char *myname;

  /* Open the scores file. */
  do_setuid ();
  stream = fopen (SCORES_FILE, "a");
  undo_setuid ();

  /* Write the score to the file. */
  if (stream)
    {
      myname = cuserid (NULL);
      if (score < 0)
        fprintf (stream, "%10s: Couldn't lift the caber.\n", myname);
      else
        fprintf (stream, "%10s: %d feet.\n", myname, score);
      fclose (stream);
      return 0;
    }
  else
    return -1;
}

29.10 Tips for Writing Setuid Programs

It is easy for setuid programs to give the user access that isn't intended—in fact, if you want to avoid this, you need to be careful. Here are some guidelines for preventing unintended access and minimizing its consequences when it does occur:

• Don't have setuid programs with privileged user IDs such as root unless it is absolutely necessary. If the resource is specific to your particular program, it's better to define a new, nonprivileged user ID or group ID just to manage that resource. It's better if you can write your program to use a special group than a special user.
• Be cautious about using the exec functions in combination with changing the effective user ID. Don't let users of your program execute arbitrary programs under a changed user ID. Executing a shell is especially bad news. Less obviously, the execlp and execvp functions are a potential risk (since the program they execute depends on the user's PATH environment variable).

  If you must exec another program under a changed ID, specify an absolute file name (see section File Name Resolution) for the executable, and make sure that the protections on that executable and all containing directories are such that ordinary users cannot replace it with some other program.

  You should also check the arguments passed to the program to make sure they do not have unexpected effects. Likewise, you should examine the environment variables. Decide which arguments and variables are safe, and reject all others.

  You should never use system in a privileged program, because it invokes a shell.

• Only use the user ID controlling the resource in the part of the program that actually uses that resource. When you're finished with it, restore the effective user ID back to the actual user's user ID. See section Enabling and Disabling Setuid Access.
• If the setuid part of your program needs to access other files besides the controlled resource, it should verify that the real user would ordinarily have permission to access those files. You can use the access function (see section How Your Access to a File is Decided) to check this; it uses the real user and group IDs, rather than the effective IDs.

29.11 Identifying Who Logged In

You can use the functions listed in this section to determine the login name of the user who is running a process, and the name of the user who logged in the current session. See also the function getuid and friends (see section Reading the Persona of a Process). How this information is collected by the system and how to control/add/remove information from the background storage is described in The User Accounting Database.

The getlogin function is declared in ‘unistd.h’, while cuserid and L_cuserid are declared in ‘stdio.h’.

Function: char * getlogin (void)

The getlogin function returns a pointer to a string containing the name of the user logged in on the controlling terminal of the process, or a null pointer if this information cannot be determined. The string is statically allocated and might be overwritten on subsequent calls to this function or to cuserid.

Function: char * cuserid (char *string)

The cuserid function returns a pointer to a string containing a user name associated with the effective ID of the process. If string is not a null pointer, it should be an array that can hold at least L_cuserid characters; the string is returned in this array. Otherwise, a pointer to a string in a static area is returned. This string is statically allocated and might be overwritten on subsequent calls to this function or to getlogin.

The use of this function is deprecated since it is marked to be withdrawn in XPG4.2 and has already been removed from newer revisions of POSIX.1.

Macro: int L_cuserid

An integer constant that indicates how long an array you might need to store a user name.

These functions let your program identify positively the user who is running or the user who logged in this session. (These can differ when setuid programs are involved; see The Persona of a Process.) The user cannot do anything to fool these functions.

For most purposes, it is more useful to use the environment variable LOGNAME to find out who the user is. This is more flexible precisely because the user can set LOGNAME arbitrarily. See section Standard Environment Variables.

29.12 The User Accounting Database

Most Unix-like operating systems keep track of logged in users by maintaining a user accounting database. This user accounting database stores for each terminal, who has logged on, at what time, the process ID of the user's login shell, etc., etc., but also stores information about the run level of the system, the time of the last system reboot, and possibly more.

The user accounting database typically lives in ‘/etc/utmp’, ‘/var/adm/utmp’ or ‘/var/run/utmp’. However, these files should never be accessed directly. For reading information from and writing information to the user accounting database, the functions described in this section should be used.

29.12.1 Manipulating the User Accounting Database

These functions and the corresponding data structures are declared in the header file ‘utmp.h’.

Data Type: struct exit_status

The exit_status data structure is used to hold information about the exit status of processes marked as DEAD_PROCESS in the user accounting database.

short int e_termination

The exit status of the process.

short int e_exit

The exit status of the process.

Data Type: struct utmp

The utmp data structure is used to hold information about entries in the user accounting database. On the GNU system it has the following members:

short int ut_type

Specifies the type of login; one of EMPTY, RUN_LVL, BOOT_TIME, OLD_TIME, NEW_TIME, INIT_PROCESS, LOGIN_PROCESS, USER_PROCESS, DEAD_PROCESS or ACCOUNTING.

pid_t ut_pid

The process ID number of the login process.

char ut_line[]

The device name of the tty (without ‘/dev/’).

char ut_id[]

The inittab ID of the process.

char ut_user[]

The user's login name.

char ut_host[]

The name of the host from which the user logged in.

struct exit_status ut_exit

The exit status of a process marked as DEAD_PROCESS.

long ut_session

The Session ID, used for windowing.

struct timeval ut_tv

Time the entry was made. For entries of type OLD_TIME this is the time when the system clock changed, and for entries of type NEW_TIME this is the time the system clock was set to.

int32_t ut_addr_v6[4]

The Internet address of a remote host.

The ut_type, ut_pid, ut_id, ut_tv, and ut_host fields are not available on all systems. Portable applications therefore should be prepared for these situations. To help doing this the ‘utmp.h’ header provides macros _HAVE_UT_TYPE, _HAVE_UT_PID, _HAVE_UT_ID, _HAVE_UT_TV, and _HAVE_UT_HOST if the respective field is available. The programmer can handle the situations by using #ifdef in the program code.

The following macros are defined for use as values for the ut_type member of the utmp structure. The values are integer constants.

EMPTY

This macro is used to indicate that the entry contains no valid user accounting information.

RUN_LVL

This macro is used to identify the systems runlevel.

BOOT_TIME

This macro is used to identify the time of system boot.

OLD_TIME

This macro is used to identify the time when the system clock changed.

NEW_TIME

This macro is used to identify the time after the system changed.

INIT_PROCESS

This macro is used to identify a process spawned by the init process.

LOGIN_PROCESS

This macro is used to identify the session leader of a logged in user.

USER_PROCESS

This macro is used to identify a user process.

DEAD_PROCESS

This macro is used to identify a terminated process.

ACCOUNTING

???

The size of the ut_line, ut_id, ut_user and ut_host arrays can be found using the sizeof operator.

Many older systems have, instead of an ut_tv member, an ut_time member, usually of type time_t, for representing the time associated with the entry. Therefore, for backwards compatibility only, ‘utmp.h’ defines ut_time as an alias for ut_tv.tv_sec.

Function: void setutent (void)

This function opens the user accounting database to begin scanning it. You can then call getutent, getutid or getutline to read entries and pututline to write entries.

If the database is already open, it resets the input to the beginning of the database.

Function: struct utmp * getutent (void)

The getutent function reads the next entry from the user accounting database. It returns a pointer to the entry, which is statically allocated and may be overwritten by subsequent calls to getutent. You must copy the contents of the structure if you wish to save the information or you can use the getutent_r function which stores the data in a user-provided buffer.

A null pointer is returned in case no further entry is available.

Function: void endutent (void)

This function closes the user accounting database.

Function: struct utmp * getutid (const struct utmp *id)

This function searches forward from the current point in the database for an entry that matches id. If the ut_type member of the id structure is one of RUN_LVL, BOOT_TIME, OLD_TIME or NEW_TIME the entries match if the ut_type members are identical. If the ut_type member of the id structure is INIT_PROCESS, LOGIN_PROCESS, USER_PROCESS or DEAD_PROCESS, the entries match if the ut_type member of the entry read from the database is one of these four, and the ut_id members match. However if the ut_id member of either the id structure or the entry read from the database is empty it checks if the ut_line members match instead. If a matching entry is found, getutid returns a pointer to the entry, which is statically allocated, and may be overwritten by a subsequent call to getutent, getutid or getutline. You must copy the contents of the structure if you wish to save the information.

A null pointer is returned in case the end of the database is reached without a match.

The getutid function may cache the last read entry. Therefore, if you are using getutid to search for multiple occurrences, it is necessary to zero out the static data after each call. Otherwise getutid could just return a pointer to the same entry over and over again.

Function: struct utmp * getutline (const struct utmp *line)

This function searches forward from the current point in the database until it finds an entry whose ut_type value is LOGIN_PROCESS or USER_PROCESS, and whose ut_line member matches the ut_line member of the line structure. If it finds such an entry, it returns a pointer to the entry which is statically allocated, and may be overwritten by a subsequent call to getutent, getutid or getutline. You must copy the contents of the structure if you wish to save the information.

A null pointer is returned in case the end of the database is reached without a match.

The getutline function may cache the last read entry. Therefore if you are using getutline to search for multiple occurrences, it is necessary to zero out the static data after each call. Otherwise getutline could just return a pointer to the same entry over and over again.

Function: struct utmp * pututline (const struct utmp *utmp)

The pututline function inserts the entry *utmp at the appropriate place in the user accounting database. If it finds that it is not already at the correct place in the database, it uses getutid to search for the position to insert the entry, however this will not modify the static structure returned by getutent, getutid and getutline. If this search fails, the entry is appended to the database.

The pututline function returns a pointer to a copy of the entry inserted in the user accounting database, or a null pointer if the entry could not be added. The following errno error conditions are defined for this function:

EPERM

The process does not have the appropriate privileges; you cannot modify the user accounting database.

All the get* functions mentioned before store the information they return in a static buffer. This can be a problem in multi-threaded programs since the data returned for the request is overwritten by the return value data in another thread. Therefore the GNU C Library provides as extensions three more functions which return the data in a user-provided buffer.

Function: int getutent_r (struct utmp *buffer, struct utmp **result)

The getutent_r is equivalent to the getutent function. It returns the next entry from the database. But instead of storing the information in a static buffer it stores it in the buffer pointed to by the parameter buffer.

If the call was successful, the function returns 0 and the pointer variable pointed to by the parameter result contains a pointer to the buffer which contains the result (this is most probably the same value as buffer). If something went wrong during the execution of getutent_r the function returns -1.

This function is a GNU extension.

Function: int getutid_r (const struct utmp *id, struct utmp *buffer, struct utmp **result)

This function retrieves just like getutid the next entry matching the information stored in id. But the result is stored in the buffer pointed to by the parameter buffer.

If successful the function returns 0 and the pointer variable pointed to by the parameter result contains a pointer to the buffer with the result (probably the same as result. If not successful the function return -1.

This function is a GNU extension.

Function: int getutline_r (const struct utmp *line, struct utmp *buffer, struct utmp **result)

This function retrieves just like getutline the next entry matching the information stored in line. But the result is stored in the buffer pointed to by the parameter buffer.

If successful the function returns 0 and the pointer variable pointed to by the parameter result contains a pointer to the buffer with the result (probably the same as result. If not successful the function return -1.

This function is a GNU extension.

In addition to the user accounting database, most systems keep a number of similar databases. For example most systems keep a log file with all previous logins (usually in ‘/etc/wtmp’ or ‘/var/log/wtmp’).

For specifying which database to examine, the following function should be used.

Function: int utmpname (const char *file)

The utmpname function changes the name of the database to be examined to file, and closes any previously opened database. By default getutent, getutid, getutline and pututline read from and write to the user accounting database.

The following macros are defined for use as the file argument:

Macro: char * _PATH_UTMP

This macro is used to specify the user accounting database.

Macro: char * _PATH_WTMP

This macro is used to specify the user accounting log file.

The utmpname function returns a value of 0 if the new name was successfully stored, and a value of -1 to indicate an error. Note that utmpname does not try to open the database, and that therefore the return value does not say anything about whether the database can be successfully opened.

Specially for maintaining log-like databases the GNU C Library provides the following function:

Function: void updwtmp (const char *wtmp_file, const struct utmp *utmp)

The updwtmp function appends the entry *utmp to the database specified by wtmp_file. For possible values for the wtmp_file argument see the utmpname function.

Portability Note: Although many operating systems provide a subset of these functions, they are not standardized. There are often subtle differences in the return types, and there are considerable differences between the various definitions of struct utmp. When programming for the GNU system, it is probably best to stick with the functions described in this section. If however, you want your program to be portable, consider using the XPG functions described in XPG User Accounting Database Functions, or take a look at the BSD compatible functions in Logging In and Out.

29.12.2 XPG User Accounting Database Functions

These functions, described in the X/Open Portability Guide, are declared in the header file ‘utmpx.h’.

Data Type: struct utmpx

The utmpx data structure contains at least the following members:

short int ut_type

Specifies the type of login; one of EMPTY, RUN_LVL, BOOT_TIME, OLD_TIME, NEW_TIME, INIT_PROCESS, LOGIN_PROCESS, USER_PROCESS or DEAD_PROCESS.

pid_t ut_pid

The process ID number of the login process.

char ut_line[]

The device name of the tty (without ‘/dev/’).

char ut_id[]

The inittab ID of the process.

char ut_user[]

The user's login name.

struct timeval ut_tv

Time the entry was made. For entries of type OLD_TIME this is the time when the system clock changed, and for entries of type NEW_TIME this is the time the system clock was set to.

On the GNU system, struct utmpx is identical to struct utmp except for the fact that including ‘utmpx.h’ does not make visible the declaration of struct exit_status.

The following macros are defined for use as values for the ut_type member of the utmpx structure. The values are integer constants and are, on the GNU system, identical to the definitions in ‘utmp.h’.

EMPTY

This macro is used to indicate that the entry contains no valid user accounting information.

RUN_LVL

This macro is used to identify the systems runlevel.

BOOT_TIME

This macro is used to identify the time of system boot.

OLD_TIME

This macro is used to identify the time when the system clock changed.

NEW_TIME

This macro is used to identify the time after the system changed.

INIT_PROCESS

This macro is used to identify a process spawned by the init process.

LOGIN_PROCESS

This macro is used to identify the session leader of a logged in user.

USER_PROCESS

This macro is used to identify a user process.

DEAD_PROCESS

This macro is used to identify a terminated process.

The size of the ut_line, ut_id and ut_user arrays can be found using the sizeof operator.

Function: void setutxent (void)

This function is similar to setutent. On the GNU system it is simply an alias for setutent.

Function: struct utmpx * getutxent (void)

The getutxent function is similar to getutent, but returns a pointer to a struct utmpx instead of struct utmp. On the GNU system it simply is an alias for getutent.

Function: void endutxent (void)

This function is similar to endutent. On the GNU system it is simply an alias for endutent.

Function: struct utmpx * getutxid (const struct utmpx *id)

This function is similar to getutid, but uses struct utmpx instead of struct utmp. On the GNU system it is simply an alias for getutid.

Function: struct utmpx * getutxline (const struct utmpx *line)

This function is similar to getutid, but uses struct utmpx instead of struct utmp. On the GNU system it is simply an alias for getutline.

Function: struct utmpx * pututxline (const struct utmpx *utmp)

The pututxline function is functionally identical to pututline, but uses struct utmpx instead of struct utmp. On the GNU system, pututxline is simply an alias for pututline.

Function: int utmpxname (const char *file)

The utmpxname function is functionally identical to utmpname. On the GNU system, utmpxname is simply an alias for utmpname.

You can translate between a traditional struct utmp and an XPG struct utmpx with the following functions. On the GNU system, these functions are merely copies, since the two structures are identical.

Function: int getutmp (const struct utmpx *utmpx, struct utmp *utmp)

getutmp copies the information, insofar as the structures are compatible, from utmpx to utmp.

Function: int getutmpx (const struct utmp *utmp, struct utmpx *utmpx)

getutmpx copies the information, insofar as the structures are compatible, from utmp to utmpx.

29.12.3 Logging In and Out

These functions, derived from BSD, are available in the separate ‘libutil’ library, and declared in ‘utmp.h’.

Note that the ut_user member of struct utmp is called ut_name in BSD. Therefore, ut_name is defined as an alias for ut_user in ‘utmp.h’.

Function: int login_tty (int filedes)

This function makes filedes the controlling terminal of the current process, redirects standard input, standard output and standard error output to this terminal, and closes filedes.

This function returns 0 on successful completion, and -1 on error.

Function: void login (const struct utmp *entry)

The login functions inserts an entry into the user accounting database. The ut_line member is set to the name of the terminal on standard input. If standard input is not a terminal login uses standard output or standard error output to determine the name of the terminal. If struct utmp has a ut_type member, login sets it to USER_PROCESS, and if there is an ut_pid member, it will be set to the process ID of the current process. The remaining entries are copied from entry.

A copy of the entry is written to the user accounting log file.

Function: int logout (const char *ut_line)

This function modifies the user accounting database to indicate that the user on ut_line has logged out.

The logout function returns 1 if the entry was successfully written to the database, or 0 on error.

Function: void logwtmp (const char *ut_line, const char *ut_name, const char *ut_host)

The logwtmp function appends an entry to the user accounting log file, for the current time and the information provided in the ut_line, ut_name and ut_host arguments.

Portability Note: The BSD struct utmp only has the ut_line, ut_name, ut_host and ut_time members. Older systems do not even have the ut_host member.

29.13 User Database

This section describes how to search and scan the database of registered users. The database itself is kept in the file ‘/etc/passwd’ on most systems, but on some systems a special network server gives access to it.

29.13.1 The Data Structure that Describes a User

The functions and data structures for accessing the system user database are declared in the header file ‘pwd.h’.

Data Type: struct passwd

The passwd data structure is used to hold information about entries in the system user data base. It has at least the following members:

char *pw_name

The user's login name.

char *pw_passwd.

The encrypted password string.

uid_t pw_uid

The user ID number.

gid_t pw_gid

The user's default group ID number.

char *pw_gecos

A string typically containing the user's real name, and possibly other information such as a phone number.

char *pw_dir

The user's home directory, or initial working directory. This might be a null pointer, in which case the interpretation is system-dependent.

char *pw_shell

The user's default shell, or the initial program run when the user logs in. This might be a null pointer, indicating that the system default should be used.

29.13.2 Looking Up One User

You can search the system user database for information about a specific user using getpwuid or getpwnam. These functions are declared in ‘pwd.h’.

Function: struct passwd * getpwuid (uid_t uid)

This function returns a pointer to a statically-allocated structure containing information about the user whose user ID is uid. This structure may be overwritten on subsequent calls to getpwuid.

A null pointer value indicates there is no user in the data base with user ID uid.

Function: int getpwuid_r (uid_t uid, struct passwd *result_buf, char *buffer, size_t buflen, struct passwd **result)

This function is similar to getpwuid in that it returns information about the user whose user ID is uid. However, it fills the user supplied structure pointed to by result_buf with the information instead of using a static buffer. The first buflen bytes of the additional buffer pointed to by buffer are used to contain additional information, normally strings which are pointed to by the elements of the result structure.

If a user with ID uid is found, the pointer returned in result points to the record which contains the wanted data (i.e., result contains the value result_buf). If no user is found or if an error occurred, the pointer returned in result is a null pointer. The function returns zero or an error code. If the buffer buffer is too small to contain all the needed information, the error code ERANGE is returned and errno is set to ERANGE.

Function: struct passwd * getpwnam (const char *name)

This function returns a pointer to a statically-allocated structure containing information about the user whose user name is name. This structure may be overwritten on subsequent calls to getpwnam.

A null pointer return indicates there is no user named name.

Function: int getpwnam_r (const char *name, struct passwd *result_buf, char *buffer, size_t buflen, struct passwd **result)

This function is similar to getpwnam in that is returns information about the user whose user name is name. However, like getpwuid_r, it fills the user supplied buffers in result_buf and buffer with the information instead of using a static buffer.

The return values are the same as for getpwuid_r.

29.13.3 Scanning the List of All Users

This section explains how a program can read the list of all users in the system, one user at a time. The functions described here are declared in ‘pwd.h’.

You can use the fgetpwent function to read user entries from a particular file.

Function: struct passwd * fgetpwent (FILE *stream)

This function reads the next user entry from stream and returns a pointer to the entry. The structure is statically allocated and is rewritten on subsequent calls to fgetpwent. You must copy the contents of the structure if you wish to save the information.

The stream must correspond to a file in the same format as the standard password database file.

Function: int fgetpwent_r (FILE *stream, struct passwd *result_buf, char *buffer, size_t buflen, struct passwd **result)

This function is similar to fgetpwent in that it reads the next user entry from stream. But the result is returned in the structure pointed to by result_buf. The first buflen bytes of the additional buffer pointed to by buffer are used to contain additional information, normally strings which are pointed to by the elements of the result structure.

The stream must correspond to a file in the same format as the standard password database file.

If the function returns zero result points to the structure with the wanted data (normally this is in result_buf). If errors occurred the return value is nonzero and result contains a null pointer.

The way to scan all the entries in the user database is with setpwent, getpwent, and endpwent.

Function: void setpwent (void)

This function initializes a stream which getpwent and getpwent_r use to read the user database.

Function: struct passwd * getpwent (void)

The getpwent function reads the next entry from the stream initialized by setpwent. It returns a pointer to the entry. The structure is statically allocated and is rewritten on subsequent calls to getpwent. You must copy the contents of the structure if you wish to save the information.

A null pointer is returned when no more entries are available.

Function: int getpwent_r (struct passwd *result_buf, char *buffer, int buflen, struct passwd **result)

This function is similar to getpwent in that it returns the next entry from the stream initialized by setpwent. Like fgetpwent_r, it uses the user-supplied buffers in result_buf and buffer to return the information requested.

The return values are the same as for fgetpwent_r.

Function: void endpwent (void)

This function closes the internal stream used by getpwent or getpwent_r.

29.13.4 Writing a User Entry

Function: int putpwent (const struct passwd *p, FILE *stream)

This function writes the user entry *p to the stream stream, in the format used for the standard user database file. The return value is zero on success and nonzero on failure.

This function exists for compatibility with SVID. We recommend that you avoid using it, because it makes sense only on the assumption that the struct passwd structure has no members except the standard ones; on a system which merges the traditional Unix data base with other extended information about users, adding an entry using this function would inevitably leave out much of the important information.

The group and user ID fields are left empty if the group or user name starts with a - or +.

The function putpwent is declared in ‘pwd.h’.

29.14 Group Database

This section describes how to search and scan the database of registered groups. The database itself is kept in the file ‘/etc/group’ on most systems, but on some systems a special network service provides access to it.

29.14.1 The Data Structure for a Group

The functions and data structures for accessing the system group database are declared in the header file ‘grp.h’.

Data Type: struct group

The group structure is used to hold information about an entry in the system group database. It has at least the following members:

char *gr_name

The name of the group.

gid_t gr_gid

The group ID of the group.

char **gr_mem

A vector of pointers to the names of users in the group. Each user name is a null-terminated string, and the vector itself is terminated by a null pointer.

29.14.2 Looking Up One Group

You can search the group database for information about a specific group using getgrgid or getgrnam. These functions are declared in ‘grp.h’.

Function: struct group * getgrgid (gid_t gid)

This function returns a pointer to a statically-allocated structure containing information about the group whose group ID is gid. This structure may be overwritten by subsequent calls to getgrgid.

A null pointer indicates there is no group with ID gid.

Function: int getgrgid_r (gid_t gid, struct group *result_buf, char *buffer, size_t buflen, struct group **result)

This function is similar to getgrgid in that it returns information about the group whose group ID is gid. However, it fills the user supplied structure pointed to by result_buf with the information instead of using a static buffer. The first buflen bytes of the additional buffer pointed to by buffer are used to contain additional information, normally strings which are pointed to by the elements of the result structure.

If a group with ID gid is found, the pointer returned in result points to the record which contains the wanted data (i.e., result contains the value result_buf). If no group is found or if an error occurred, the pointer returned in result is a null pointer. The function returns zero or an error code. If the buffer buffer is too small to contain all the needed information, the error code ERANGE is returned and errno is set to ERANGE.

Function: struct group * getgrnam (const char *name)

This function returns a pointer to a statically-allocated structure containing information about the group whose group name is name. This structure may be overwritten by subsequent calls to getgrnam.

A null pointer indicates there is no group named name.

Function: int getgrnam_r (const char *name, struct group *result_buf, char *buffer, size_t buflen, struct group **result)

This function is similar to getgrnam in that is returns information about the group whose group name is name. Like getgrgid_r, it uses the user supplied buffers in result_buf and buffer, not a static buffer.

The return values are the same as for getgrgid_r ERANGE.

29.14.3 Scanning the List of All Groups

This section explains how a program can read the list of all groups in the system, one group at a time. The functions described here are declared in ‘grp.h’.

You can use the fgetgrent function to read group entries from a particular file.

Function: struct group * fgetgrent (FILE *stream)

The fgetgrent function reads the next entry from stream. It returns a pointer to the entry. The structure is statically allocated and is overwritten on subsequent calls to fgetgrent. You must copy the contents of the structure if you wish to save the information.

The stream must correspond to a file in the same format as the standard group database file.

Function: int fgetgrent_r (FILE *stream, struct group *result_buf, char *buffer, size_t buflen, struct group **result)

This function is similar to fgetgrent in that it reads the next user entry from stream. But the result is returned in the structure pointed to by result_buf. The first buflen bytes of the additional buffer pointed to by buffer are used to contain additional information, normally strings which are pointed to by the elements of the result structure.

This stream must correspond to a file in the same format as the standard group database file.

If the function returns zero result points to the structure with the wanted data (normally this is in result_buf). If errors occurred the return value is non-zero and result contains a null pointer.

The way to scan all the entries in the group database is with setgrent, getgrent, and endgrent.

Function: void setgrent (void)

This function initializes a stream for reading from the group data base. You use this stream by calling getgrent or getgrent_r.

Function: struct group * getgrent (void)

The getgrent function reads the next entry from the stream initialized by setgrent. It returns a pointer to the entry. The structure is statically allocated and is overwritten on subsequent calls to getgrent. You must copy the contents of the structure if you wish to save the information.

Function: int getgrent_r (struct group *result_buf, char *buffer, size_t buflen, struct group **result)

This function is similar to getgrent in that it returns the next entry from the stream initialized by setgrent. Like fgetgrent_r, it places the result in user-supplied buffers pointed to result_buf and buffer.

If the function returns zero result contains a pointer to the data (normally equal to result_buf). If errors occurred the return value is non-zero and result contains a null pointer.

Function: void endgrent (void)

This function closes the internal stream used by getgrent or getgrent_r.

29.15 User and Group Database Example

Here is an example program showing the use of the system database inquiry functions. The program prints some information about the user running the program.

#include <grp.h>
#include <pwd.h>
#include <sys/types.h>
#include <unistd.h>
#include <stdlib.h>

int
main (void)
{
  uid_t me;
  struct passwd *my_passwd;
  struct group *my_group;
  char **members;

  /* Get information about the user ID. */
  me = getuid ();
  my_passwd = getpwuid (me);
  if (!my_passwd)
    {
      printf ("Couldn't find out about user %d.\n", (int) me);
      exit (EXIT_FAILURE);
    }

  /* Print the information. */
  printf ("I am %s.\n", my_passwd->pw_gecos);
  printf ("My login name is %s.\n", my_passwd->pw_name);
  printf ("My uid is %d.\n", (int) (my_passwd->pw_uid));
  printf ("My home directory is %s.\n", my_passwd->pw_dir);
  printf ("My default shell is %s.\n", my_passwd->pw_shell);

  /* Get information about the default group ID. */
  my_group = getgrgid (my_passwd->pw_gid);
  if (!my_group)
    {
      printf ("Couldn't find out about group %d.\n",
              (int) my_passwd->pw_gid);
      exit (EXIT_FAILURE);
    }

  /* Print the information. */
  printf ("My default group is %s (%d).\n",
          my_group->gr_name, (int) (my_passwd->pw_gid));
  printf ("The members of this group are:\n");
  members = my_group->gr_mem;
  while (*members)
    {
      printf ("  %s\n", *(members));
      members++;
    }

  return EXIT_SUCCESS;
}

Here is some output from this program:

I am Throckmorton Snurd.
My login name is snurd.
My uid is 31093.
My home directory is /home/fsg/snurd.
My default shell is /bin/sh.
My default group is guest (12).
The members of this group are:
  friedman
  tami

29.16 Netgroup Database

29.16.1 Netgroup Data

Sometimes it is useful to group users according to other criteria (see section Group Database). E.g., it is useful to associate a certain group of users with a certain machine. On the other hand grouping of host names is not supported so far.

In Sun Microsystems SunOS appeared a new kind of database, the netgroup database. It allows grouping hosts, users, and domains freely, giving them individual names. To be more concrete, a netgroup is a list of triples consisting of a host name, a user name, and a domain name where any of the entries can be a wildcard entry matching all inputs. A last possibility is that names of other netgroups can also be given in the list specifying a netgroup. So one can construct arbitrary hierarchies without loops.

Sun's implementation allows netgroups only for the nis or nisplus service, see section Services in the NSS configuration File. The implementation in the GNU C library has no such restriction. An entry in either of the input services must have the following form:

groupname ( groupname | (hostname,username,domainname) )+

Any of the fields in the triple can be empty which means anything matches. While describing the functions we will see that the opposite case is useful as well. I.e., there may be entries which will not match any input. For entries like this, a name consisting of the single character - shall be used.

29.16.2 Looking up one Netgroup

The lookup functions for netgroups are a bit different to all other system database handling functions. Since a single netgroup can contain many entries a two-step process is needed. First a single netgroup is selected and then one can iterate over all entries in this netgroup. These functions are declared in ‘netdb.h’.

Function: int setnetgrent (const char *netgroup)

A call to this function initializes the internal state of the library to allow following calls of the getnetgrent to iterate over all entries in the netgroup with name netgroup.

When the call is successful (i.e., when a netgroup with this name exists) the return value is 1. When the return value is 0 no netgroup of this name is known or some other error occurred.

It is important to remember that there is only one single state for iterating the netgroups. Even if the programmer uses the getnetgrent_r function the result is not really reentrant since always only one single netgroup at a time can be processed. If the program needs to process more than one netgroup simultaneously she must protect this by using external locking. This problem was introduced in the original netgroups implementation in SunOS and since we must stay compatible it is not possible to change this.

Some other functions also use the netgroups state. Currently these are the innetgr function and parts of the implementation of the compat service part of the NSS implementation.

Function: int getnetgrent (char **hostp, char **userp, char **domainp)

This function returns the next unprocessed entry of the currently selected netgroup. The string pointers, in which addresses are passed in the arguments hostp, userp, and domainp, will contain after a successful call pointers to appropriate strings. If the string in the next entry is empty the pointer has the value NULL. The returned string pointers are only valid if none of the netgroup related functions are called.

The return value is 1 if the next entry was successfully read. A value of 0 means no further entries exist or internal errors occurred.

Function: int getnetgrent_r (char **hostp, char **userp, char **domainp, char *buffer, int buflen)

This function is similar to getnetgrent with only one exception: the strings the three string pointers hostp, userp, and domainp point to, are placed in the buffer of buflen bytes starting at buffer. This means the returned values are valid even after other netgroup related functions are called.

The return value is 1 if the next entry was successfully read and the buffer contains enough room to place the strings in it. 0 is returned in case no more entries are found, the buffer is too small, or internal errors occurred.

This function is a GNU extension. The original implementation in the SunOS libc does not provide this function.

Function: void endnetgrent (void)

This function frees all buffers which were allocated to process the last selected netgroup. As a result all string pointers returned by calls to getnetgrent are invalid afterwards.

29.16.3 Testing for Netgroup Membership

It is often not necessary to scan the whole netgroup since often the only interesting question is whether a given entry is part of the selected netgroup.

Function: int innetgr (const char *netgroup, const char *host, const char *user, const char *domain)

This function tests whether the triple specified by the parameters hostp, userp, and domainp is part of the netgroup netgroup. Using this function has the advantage that

1. no other netgroup function can use the global netgroup state since internal locking is used and
2. the function is implemented more efficiently than successive calls to the other set/get/endnetgrent functions.

Any of the pointers hostp, userp, and domainp can be NULL which means any value is accepted in this position. This is also true for the name - which should not match any other string otherwise.

The return value is 1 if an entry matching the given triple is found in the netgroup. The return value is 0 if the netgroup itself is not found, the netgroup does not contain the triple or internal errors occurred.

This document was generated by root on January, 9 2009 using texi2html 1.78.