SQL Module 0. Introduction The SQL module is composed of two parts: a generic SQL front-end (rlm_sql), and a series of database-dependent back-end drivers, (rlm_sql_mysql, rlm_sql_postgresql, etc.) In order to build the drivers, you MUST ALSO install the development versions of the database. That is, you must have the appropriate header files and client libraries for (say) MySQL. The rlm_sql_mysql driver is NOT a complete MySQL client implementation. Instead, it is a small 'shim' between the FreeRADIUS rlm_sql module, and the MySQL client libraries. In general, the SQL schemas mirror the layout of the 'users' file. So for configuring check items and reply items, see 'man 5 users', and the examples in the 'users' file. 1. Schema and usage The schemas are available in raddb/sql//*, where is the name of the database (mysql, postgresql, etc.) The SQL module employs two sets of check and reply item tables for processing in the authorization stage. One set of tables (radcheck and radreply) are specific to a single user. The other set of tables (radgroupcheck and radgroupreply) is used to apply check and reply items to users that are members of a certain SQL group. The usergroup table provides the list of groups each user is a member of along with a priority field to control the order in which groups are processed. When a request comes into the server and is processed by the SQL module, the flow goes something like this: 1. Search the radcheck table for any check attributes specific to the user 2. If check attributes are found, and there's a match, pull the reply items from the radreply table for this user and add them to the reply 3. Group processing then begins if any of the following conditions are met: a. The user IS NOT found in radcheck b. The user IS found in radcheck, but the check items don't match c. The user IS found in radcheck, the check items DO match AND Fall-Through is set in the radreply table d. The user IS found in radcheck, the check items DO match AND the read_groups directive is set to 'yes' 4. If groups are to be processed for this user, the first thing that is done is the list of groups this user is a member of is pulled from the usergroup table ordered by the priority field. The priority field of the usergroup table allows us to control the order in which groups are processed, so that we can emulate the ordering in the users file. This can be important in many cases. 5. For each group this user is a member of, the corresponding check items are pulled from radgroupcheck table and compared with the request. If there is a match, the reply items for this group are pulled from the radgroupreply table and applied. 6. Processing continues to the next group IF: a. There was not a match for the last group's check items OR b. Fall-Through was set in the last group's reply items (The above is exactly the same as in the users file) 7. Finally, if the user has a User-Profile attribute set or the Default Profile option is set in the sql.conf, then steps 4-6 are repeated for the groups that the profile is a member of. For any fairly complex setup, it is likely that most of the actual processing will be done in the groups. In these cases, the user entry in radcheck will be of limited use except for things like setting the user's password. So, one might have the following setup: radcheck table: joeuser Cleartext-Password := somepassword radreply table: joeuser Fall-Through = Yes radgroupcheck table: Check items for various connection scenarios radgroupreply table: reply items for the groups usergroup table: joeuser WLANgroup 1(this is the priority) joeuser PPPgroup 2 2. What NOT to do. One of the fields of the SQL schema is named 'op' This is for the 'operator' used by the attributes. e.g.: Framed-IP-Address = 1.2.3.4 ^ ATTRIBUTE ----^ ^ OP ^ VALUE If you want the server to be completely misconfigured, and to never do what you want, leave the 'op' field blank. If you want to be rudely told to RTFM, then post questions on the mailing list, asking "why doesn't my SQL configuration work when I leave the 'op' field empty?" The short answer is that with the op field empty, the server does not know what you want it to do with the attribute. Should it be added to the reply? Maybe you wanted to compare the operator to one in the request? The server simply doesn't know. So put a value in the field. The value is the string form of the operator: "=", ">=", etc. See Section 4, below, for more details. 3. Authentication versus Authorization Many people ask if they can "authenticate" users to their SQL database. The answer to this question is "You're asking the wrong question." An SQL database stores information. An SQL database is NOT an authentication server. The ONLY users who should be able to authenticate themselves to the database are the people who administer it. Most administrators do NOT want every user to be able to access the database, which means that most users will not be able to "authenticate" themselves to the database. Instead, the users will have their authorization information (name, password, configuration) stored in the database. The configuration files for FreeRADIUS contain a username and password used to authenticate FreeRADIUS to the SQL server. (See raddb/sql.conf). Once the FreeRADIUS authentication server is connected to the SQL database server, then FreeRADIUS can pull user names and passwords out of the database, and use that information to perform the authentication. 4. Operators The list of operators is given below. Op Example and documentation -- ------------------------- = "Attribute = Value" Not allowed as a check item for RADIUS protocol attributes. It is allowed for server configuration attributes (Auth-Type, etc), and sets the value of on attribute, only if there is no other item of the same attribute. As a reply item, it means "add the item to the reply list, but only if there is no other item of the same attribute." := "Attribute := Value" Always matches as a check item, and replaces in the configuration items any attribute of the same name. If no attribute of that name appears in the request, then this attribute is added. As a reply item, it has an identical meaning, but for the reply items, instead of the request items. == "Attribute == Value" As a check item, it matches if the named attribute is present in the request, AND has the given value. Not allowed as a reply item. += "Attribute += Value" Always matches as a check item, and adds the current attribute with value to the list of configuration items. As a reply item, it has an identical meaning, but the attribute is added to the reply items. != "Attribute != Value" As a check item, matches if the given attribute is in the request, AND does not have the given value. Not allowed as a reply item. > "Attribute > Value" As a check item, it matches if the request contains an attribute with a value greater than the one given. Not allowed as a reply item. >= "Attribute >= Value" As a check item, it matches if the request contains an attribute with a value greater than, or equal to the one given. Not allowed as a reply item. < "Attribute < Value" As a check item, it matches if the request contains an attribute with a value less than the one given. Not allowed as a reply item. <= "Attribute <= Value" As a check item, it matches if the request contains an attribute with a value less than, or equal to the one given. Not allowed as a reply item. =~ "Attribute =~ Expression" As a check item, it matches if the request contains an attribute which matches the given regular expression. This operator may only be applied to string attributes. Not allowed as a reply item. !~ "Attribute !~ Expression" As a check item, it matches if the request contains an attribute which does not match the given regular expression. This operator may only be applied to string attributes. Not allowed as a reply item. =* "Attribute =* Value" As a check item, it matches if the request contains the named attribute, no matter what the value is. Not allowed as a reply item. !* "Attribute !* Value" As a check item, it matches if the request does not contain the named attribute, no matter what the value is. Not allowed as a reply item. 5. Instances Just like any other module, multiple instances of the rlm_sql module can be defined and used wherever you like. The default .conf files for the different database types, contain 1 instance without a name like so: sql { ... } You can create multiple named instances like so: sql sql_instance1 { ... } sql sql_instance2 { ... } And then you can use a specific instance in radiusd.conf, like so: authorize { ... sql_instance1 ... } accounting { ... sql_instance1 sql_instance2 ... }