Authorizations

An authorization is a unique string that represents a user's right to perform some operation or class of operations. Authorization definitions are stored in a database called auth_attr(5). For programming authorization checks, only the authorization name is significant.

Some typical values in an auth_attr database are shown below.

solaris.jobs.:::Cron and At Jobs::help=JobHeader.html
solaris.jobs.grant:::Delegate Cron & At \


    Administration::help=JobsGrant.html
solaris.jobs.admin:::Manage All Jobs::help=AuthJobsAdmin.html
solaris.jobs.user:::Cron & At User::help=JobsUser.html

Authorization name strings ending with the grant suffix are special authorizations that give a user the ability to delegate authorizations with the same prefix and functional area to other users.

Creating Authorization Checks

To check authorizations, use the chkauthattr(3SECDB) library function, which verifies whether or not a user has a given authorization. The synopsis is:

int chkauthattr(const char *authname, const char *username);

The chkauthattr() function checks the policy.conf(5), user_attr(5), and prof_attr(5) databases in order for a match to the given authorization.

If you are modifying existing code that tests for root UID, you should find the test in the code and replace it with the chkauthattr() function. A typical root UID check is shown in the first code segment below. An authorization check replacing it is shown in the second code segment; it uses the solaris.jobs.admin authorization and a variable called real_login representing the user.

Example 1 Standard root check

ruid = getuid();
if ((eflag || lflag || rflag) && argc == 1) {


        if ((pwp = getpwnam(*argv)) == NULL)


                crabort(INVALIDUSER);


        if (ruid != 0) {


                if (pwp->pw_uid != ruid)


                        crabort(NOTROOT);


                else


                        pp = getuser(ruid);


        } else


                pp = *argv++;
} else {

Example 2 Authorization check

ruid = getuid();
if ((pwp = getpwuid(ruid)) == NULL)


        crabort(INVALIDUSER);
strcpy(real_login, pwp->pw_name);
if ((eflag || lflag || rflag) && argc == 1) {


        if ((pwp = getpwnam(*argv)) == NULL)


                crabort(INVALIDUSER);


        if (!chkauthattr("solaris.jobs.admin", real_login)) {


                if (pwp->pw_uid != ruid)


                        crabort(NOTROOT);


                else


                        pp = getuser(ruid);


        } else


                pp = *argv++;
} else {

For new applications, find an appropriate location for the test and use chkauthattr() as shown above. Typically the authorization check makes an access decision based on the identity of the calling user to determine if a privileged action (for example, a system call) should be taken on behalf of that user.

Applications that perform a test to restrict who can perform their security-relevant functionality are generally setuid to root. Programs that were written prior to RBAC and that are only available to the root user may not have such checks. In most cases, the kernel requires an effective user ID of root to override policy enforcement. Therefore, authorization checking is most useful in programs that are setuid to root.

For instance, if you want to write a program that allows authorized users to set the system date, the command must be run with an effective user ID of root. Typically, this means that the file modes for the file would be -rwsr-xr-x with root ownership.

Use caution, though, when making programs setuid to root. For example, the effective UID should be set to the real UID as early as possible in the program's initialization function. The effective UID can then be set back to root after the authorization check is performed and before the system call is made. On return from the system call, the effective UID should be set back to the real UID again to adhere to the principle of least privilege.

Another consideration is that LD_LIBRARY path is ignored for setuid programs (see SECURITY section in ld.so.1(1)) and that shell scripts must be modified to work properly when the effective and real UIDs are different. For example, the -p flag in Bourne shell is required to avoid resetting the effective UID back to the real UID.

Using an effective UID of root instead of the real UID requires extra care when writing shell scripts. For example, many shell scripts check to see if the user is root before executing their functionality. With RBAC, these shell scripts may be running with the effective UID of root and with a real UID of a user or role. Thus, the shell script should check euid instead of uid. For example,

WHO=`id | cut -f1 -d" "`
if [ ! "$WHO" = "uid=0(root)" ]
then


       echo "$PROG: ERROR: you must be super-user to run this script."


       exit 1
fi

should be changed to

WHO=`/usr/xpg4/bin/id -n -u`
if [ ! "$WHO" = "root" ]
then


       echo "$PROG: ERROR: you are not authorized to run this script."


       exit 1
fi

Authorizations can be explicitly checked in shell scripts by checking the output of the auths(1) utility. For example,

for auth in `auths | tr , " "` NOTFOUND
do


    [ "$auth" = "solaris.date" ] && break       # authorization found
done
if [ "$auth" != "solaris.date" ]
then


    echo >&2 "$PROG: ERROR: you are not authorized to set the date"


    exit 1
fi