Langue: en

Version: 2007/09/04 (mandriva - 22/10/07)

Section: 5 (Format de fichier)


slapd.access - access configuration for slapd, the stand-alone LDAP daemon




The slapd.conf(5) file contains configuration information for the slapd(8) daemon. This configuration file is also used by the SLAPD tools slapacl(8), slapadd(8), slapauth(8), slapcat(8), slapdn(8), slapindex(8), and slaptest(8).

The slapd.conf file consists of a series of global configuration options that apply to slapd as a whole (including all backends), followed by zero or more database backend definitions that contain information specific to a backend instance.

The general format of slapd.conf is as follows:

     # comment - these options apply to every database
     <global configuration options>
     # first database definition & configuration options
     database    <backend 1 type>
     <configuration options specific to backend 1>
     # subsequent database definitions & configuration options

Both the global configuration and each backend-specific section can contain access information. Backend-specific access control directives are used for those entries that belong to the backend, according to their naming context. In case no access control directives are defined for a backend or those which are defined are not applicable, the directives from the global configuration section are then used.

If no access controls are present, the default policy allows anyone and everyone to read anything but restricts updates to rootdn. (e.g., "access to * by * read"). The rootdn can always read and write EVERYTHING!

For entries not held in any backend (such as a root DSE), the directives of the first backend (and any global directives) are used.

Arguments that should be replaced by actual text are shown in brackets <>.


The structure of the access control directives is
access to <what> [ by <who> [ <access> ] [ <control> ] ]+
Grant access (specified by <access>) to a set of entries and/or attributes (specified by <what>) by one or more requestors (specified by <who>).

Lists of access directives are evaluated in the order they appear in slapd.conf. When a <what> clause matches the datum whose access is being evaluated, its <who> clause list is checked. When a <who> clause matches the accessor's properties, its <access> and <control> clauses are evaluated. Access control checking stops at the first match of the <what> and <who> clause, unless otherwise dictated by the <control> clause. Each <who> clause list is implicitly terminated by a

         by * none stop

clause that results in stopping the access control with no access privileges granted. Each <what> clause list is implicitly terminated by a

         access to *
                 by * none

clause that results in granting no access privileges to an otherwise unspecified datum.


The field <what> specifies the entity the access control directive applies to. It can have the forms
         attrs=<attrlist>[ val[/matchingRule][.<attrstyle>]=<attrval>]



The statement dn=<dnpattern> selects the entries based on their naming context. The <dnpattern> is a string representation of the entry's DN. The wildcard * stands for all the entries, and it is implied if no dn form is given.

The <dnstyle> is optional; however, it is recommended to specify it to avoid ambiguities. Base (synonym of baseObject), the default, or exact (an alias of base) indicates the entry whose DN is equal to the <dnpattern>; one (synonym of onelevel) indicates all the entries immediately below the <dnpattern>, sub (synonym of subtree) indicates all entries in the subtree at the <dnpattern>, children indicates all the entries below (subordinate to) the <dnpattern>.

If the <dnstyle> qualifier is regex, then <dnpattern> is a POSIX (''extended'') regular expression pattern, as detailed in regex(7) and/or re_format(7), matching a normalized string representation of the entry's DN. The regex form of the pattern does not (yet) support UTF-8.

The statement filter=<ldapfilter> selects the entries based on a valid LDAP filter as described in RFC 4515. A filter of (objectClass=*) is implied if no filter form is given.

The statement attrs=<attrlist> selects the attributes the access control rule applies to. It is a comma-separated list of attribute types, plus the special names entry, indicating access to the entry itself, and children, indicating access to the entry's children. ObjectClass names may also be specified in this list, which will affect all the attributes that are required and/or allowed by that objectClass. Actually, names in <attrlist> that are prefixed by @ are directly treated as objectClass names. A name prefixed by ! is also treated as an objectClass, but in this case the access rule affects the attributes that are not required nor allowed by that objectClass. If no attrs form is given, attrs=@extensibleObject is implied, i.e. all attributes are addressed.

Using the form attrs=<attr> val[/matchingRule][.<attrstyle>]=<attrval> specifies access to a particular value of a single attribute. In this case, only a single attribute type may be given. The <attrstyle> exact (the default) uses the attribute's equality matching rule to compare the value, unless a different (and compatible) matching rule is specified. If the <attrstyle> is regex, the provided value is used as a POSIX (''extended'') regular expression pattern. If the attribute has DN syntax, the <attrstyle> can be any of base, onelevel, subtree or children, resulting in base, onelevel, subtree or children match, respectively.

The dn, filter, and attrs statements are additive; they can be used in sequence to select entities the access rule applies to based on naming context, value and attribute type simultaneously.


The field <who> indicates whom the access rules apply to. Multiple <who> statements can appear in an access control statement, indicating the different access privileges to the same resource that apply to different accessee. It can have the forms


         <name>=aci              <pattern>=<attrname>]

They may be specified in combination.


The wildcard * refers to everybody.

The keywords prefixed by real act as their counterparts without prefix; the checking respectively occurs with the authentication DN and the authorization DN.

The keyword anonymous means access is granted to unauthenticated clients; it is mostly used to limit access to authentication resources (e.g. the userPassword attribute) to unauthenticated clients for authentication purposes.

The keyword users means access is granted to authenticated clients.

The keyword self means access to an entry is allowed to the entry itself (e.g. the entry being accessed and the requesting entry must be the same). It allows the level{<n>} style, where <n> indicates what ancestor of the DN is to be used in matches. A positive value indicates that the <n>-th ancestor of the user's DN is to be considered; a negative value indicates that the <n>-th ancestor of the target is to be considered. For example, a "by self.level{1} ..." clause would match when the object "dc=example,dc=com" is accessed by "cn=User,dc=example,dc=com". A "by self.level{-1} ..." clause would match when the same user accesses the object "ou=Address Book,cn=User,dc=example,dc=com".

The statement dn=<DN> means that access is granted to the matching DN. The optional style qualifier dnstyle allows the same choices of the dn form of the <what> field. In addition, the regex style can exploit substring substitution of submatches in the <what> dn.regex clause by using the form $<digit>, with digit ranging from 0 to 9 (where 0 matches the entire string), or the form ${<digit>+}, for submatches higher than 9. Since the dollar character is used to indicate a substring replacement, the dollar character that is used to indicate match up to the end of the string must be escaped by a second dollar character, e.g.

     access to dn.regex="^(.+,)?uid=([^,]+),dc=[^,]+,dc=com$"
         by dn.regex="^uid=$2,dc=[^,]+,dc=com$$" write

The style qualifier allows an optional modifier. At present, the only type allowed is expand, which causes substring substitution of submatches to take place even if dnstyle is not regex. Note that the regex dnstyle in the above example may be of use only if the <by> clause needs to be a regex; otherwise, if the value of the second (from the right) dc= portion of the DN in the above example were fixed, the form

     access to dn.regex="^(.+,)?uid=([^,]+),dc=example,dc=com$"
         by dn.exact,expand="uid=$2,dc=example,dc=com" write

could be used; if it had to match the value in the <what> clause, the form

     access to dn.regex="^(.+,)?uid=([^,]+),dc=([^,]+),dc=com$"
         by dn.exact,expand="uid=$2,dc=$3,dc=com" write

could be used.

Forms of the <what> clause other than regex may provide submatches as well. The base(object), the sub(tree), the one(level), and the children forms provide $0 as the match of the entire string. The sub(tree), the one(level), and the children forms also provide $1 as the match of the rightmost part of the DN as defined in the <what> clause. This may be useful, for instance, to provide access to all the ancestors of a user by defining

     access to dn.subtree="dc=com"
         by dn.subtree,expand="$1" read

which means that only access to entries that appear in the DN of the <by> clause is allowed.

The level{<n>} form is an extension and a generalization of the onelevel form, which matches all DNs whose <n>-th ancestor is the pattern. So, level{1} is equivalent to onelevel, and level{0} is equivalent to base.

It is perfectly useless to give any access privileges to a DN that exactly matches the rootdn of the database the ACLs apply to, because it implicitly possesses write privileges for the entire tree of that database. Actually, access control is bypassed for the rootdn, to solve the intrinsic chicken-and-egg problem.

The statement dnattr=<attrname> means that access is granted to requests whose DN is listed in the entry being accessed under the <attrname> attribute.

The statement group=<group> means that access is granted to requests whose DN is listed in the group entry whose DN is given by <group>. The optional parameters <objectclass> and <attrname> define the objectClass and the member attributeType of the group entry. The defaults are groupOfNames and member, respectively. The optional style qualifier <style> can be expand, which means that <group> will be expanded as a replacement string (but not as a regular expression) according to regex(7) and/or re_format(7), and exact, which means that exact match will be used. If the style of the DN portion of the <what> clause is regex, the submatches are made available according to regex(7) and/or re_format(7); other styles provide limited submatches as discussed above about the DN form of the <by> clause.

For static groups, the specified attributeType must have DistinguishedName or NameAndOptionalUID syntax. For dynamic groups the attributeType must be a subtype of the labeledURI attributeType. Only LDAP URIs of the form ldap:///<base>??<scope>?<filter> will be evaluated in a dynamic group, by searching the local server only.

The statements peername=<peername>, sockname=<sockname>, domain=<domain>, and sockurl=<sockurl> mean that the contacting host IP (in the form IP=<ip>:<port> for IPv4, or IP=[<ipv6>]:<port> for IPv6) or the contacting host named pipe file name (in the form PATH=<path> if connecting through a named pipe) for peername, the named pipe file name for sockname, the contacting host name for domain, and the contacting URL for sockurl are compared against pattern to determine access. The same style rules for pattern match described for the group case apply, plus the regex style, which implies submatch expand and regex match of the corresponding connection parameters. The exact style of the <peername> clause (the default) implies a case-exact match on the client's IP, including the IP= prefix and the trailing :<port>, or the client's path, including the PATH= prefix if connecting through a named pipe. The special ip style interprets the pattern as <peername>=<ip>[%<mask>][{<n>}], where <ip> and <mask> are dotted digit representations of the IP and the mask, while <n>, delimited by curly brackets, is an optional port. The same applies to IPv6 addresses when the special ipv6 style is used. When checking access privileges, the IP portion of the peername is extracted, eliminating the IP= prefix and the :<port> part, and it is compared against the <ip> portion of the pattern after masking with <mask>. As an example, peername.ip= and peername.ipv6=::1 allow connections only from localhost, peername.ip= allows connections from any IP in the 192.168.1 class C domain, and peername.ip={9009} allows connections from any IP in the 192.168.1.[16-31] range of the same domain, only if port 9009 is used. The special path style eliminates the PATH= prefix from the peername when connecting through a named pipe, and performs an exact match on the given pattern. The <domain> clause also allows the subtree style, which succeeds when a fully qualified name exactly matches the domain pattern, or its trailing part, after a dot, exactly matches the domain pattern. The expand style is allowed, implying an exact match with submatch expansion; the use of expand as a style modifier is considered more appropriate. As an example, will match, but will not match The domain of the contacting host is determined by performing a DNS reverse lookup. As this lookup can easily be spoofed, use of the domain statement is strongly discouraged. By default, reverse lookups are disabled. The optional domainstyle qualifier of the <domain> clause allows a modifier option; the only value currently supported is expand, which causes substring substitution of submatches to take place even if the domainstyle is not regex, much like the analogous usage in <dn> clause.

The statement set=<pattern> is undocumented yet.

The statement dynacl/<name>[/<options>][.<dynstyle>][=<pattern>] means that access checking is delegated to the admin-defined method indicated by <name>, which can be registered at run-time by means of the moduleload statement. The fields <options>, <dynstyle> and <pattern> are optional, and are directly passed to the registered parsing routine. Dynacl is experimental; it must be enabled at compile time.

The statement dynacl/aci[=<attrname>] means that the access control is determined by the values in the attrname of the entry itself. The optional <attrname> indicates what attributeType holds the ACI information in the entry. By default, the OpenLDAPaci operational attribute is used. ACIs are experimental; they must be enabled at compile time.

The statements ssf=<n>, transport_ssf=<n>, tls_ssf=<n>, and sasl_ssf=<n> set the minimum required Security Strength Factor (ssf) needed to grant access. The value should be positive integer.


The optional field <access> ::= [[real]self]{<level>|<priv>} determines the access level or the specific access privileges the who field will have. Its component are defined as
         <level> ::= none|disclose|auth|compare|search|read|write|manage
         <priv> ::= {=|+|-}{m|w|r|s|c|x|d|0}+

The modifier self allows special operations like having a certain access level or privilege only in case the operation involves the name of the user that's requesting the access. It implies the user that requests access is authorized. The modifier realself refers to the authenticated DN as opposed to the authorized DN of the self modifier. An example is the selfwrite access to the member attribute of a group, which allows one to add/delete its own DN from the member list of a group, without affecting other members.

The level access model relies on an incremental interpretation of the access privileges. The possible levels are none, disclose, auth, compare, search, read, and write. Each access level implies all the preceding ones, thus manage grants all access including administrative access,

The none access level disallows all access including disclosure on error.

The disclose access level allows disclosure of information on error.

The auth access level means that one is allowed access to an attribute to perform authentication/authorization operations (e.g. bind) with no other access. This is useful to grant unauthenticated clients the least possible access level to critical resources, like passwords.

The priv access model relies on the explicit setting of access privileges for each clause. The = sign resets previously defined accesses; as a consequence, the final access privileges will be only those defined by the clause. The + and - signs add/remove access privileges to the existing ones. The privileges are m for manage, w for write, r for read, s for search, c for compare, x for authentication, and d for disclose. More than one of the above privileges can be added in one statement. 0 indicates no privileges and is used only by itself (e.g., +0).

If no access is given, it defaults to +0.


The optional field <control> controls the flow of access rule application. It can have the forms

where stop, the default, means access checking stops in case of match. The other two forms are used to keep on processing access clauses. In detail, the continue form allows for other <who> clauses in the same <access> clause to be considered, so that they may result in incrementally altering the privileges, while the break form allows for other <access> clauses that match the same target to be processed. Consider the (silly) example

         access to dn.subtree="dc=example,dc=com" attrs=cn
                 by * =cs break
         access to dn.subtree="ou=People,dc=example,dc=com"
                 by * +r

which allows search and compare privileges to everybody under the "dc=example,dc=com" tree, with the second rule allowing also read in the "ou=People" subtree, or the (even more silly) example

         access to dn.subtree="dc=example,dc=com" attrs=cn
                 by * =cs continue
                 by users +r

which grants everybody search and compare privileges, and adds read privileges to authenticated clients.

One useful application is to easily grant write privileges to an updatedn that is different from the rootdn. In this case, since the updatedn needs write access to (almost) all data, one can use

         access to *
                 by dn.exact="cn=The Update DN,dc=example,dc=com" write
                 by * break

as the first access rule. As a consequence, unless the operation is performed with the updatedn identity, control is passed straight to the subsequent rules.


Operations require different privileges on different portions of entries. The following summary applies to primary database backends such as the BDB and HDB backends. Requirements for other backends may (and often do) differ.

The add operation requires write (=w) privileges on the pseudo-attribute entry of the entry being added, and write (=w) privileges on the pseudo-attribute children of the entry's parent. When adding the suffix entry of a database, write access to children of the empty DN ("") is required.

The bind operation, when credentials are stored in the directory, requires auth (=x) privileges on the attribute the credentials are stored in (usually userPassword).

The compare operation requires compare (=c) privileges on the attribute that is being compared.

The delete operation requires write (=w) privileges on the pseudo-attribute entry of the entry being deleted, and write (=w) privileges on the children pseudo-attribute of the entry's parent.

The modify operation requires write (=w) privileges on the attributes being modified.

The modrdn operation requires write (=w) privileges on the pseudo-attribute entry of the entry whose relative DN is being modified, write (=w) privileges on the pseudo-attribute children of the old and new entry's parents, and write (=w) privileges on the attributes that are present in the new relative DN. Write (=w) privileges are also required on the attributes that are present in the old relative DN if deleteoldrdn is set to 1.

The search operation, requires search (=s) privileges on the entry pseudo-attribute of the searchBase (NOTE: this was introduced with 2.3). Then, for each entry, it requires search (=s) privileges on the attributes that are defined in the filter. The resulting entries are finally tested for read (=r) privileges on the pseudo-attribute entry (for read access to the entry itself) and for read (=r) access on each value of each attribute that is requested. Also, for each referral object used in generating continuation references, the operation requires read (=r) access on the pseudo-attribute entry (for read access to the referral object itself), as well as read (=r) access to the attribute holding the referral information (generally the ref attribute).

Some internal operations and some controls require specific access privileges. The authzID mapping and the proxyAuthz control require auth (=x) privileges on all the attributes that are present in the search filter of the URI regexp maps (the right-hand side of the authz-regexp directives). Auth (=x) privileges are also required on the authzTo attribute of the authorizing identity and/or on the authzFrom attribute of the authorized identity.

Access control to search entries is checked by the frontend, so it is fully honored by all backends; for all other operations and for the discovery phase of the search operation, full ACL semantics is only supported by the primary backends, i.e. back-bdb(5), and back-hdb(5).

Some other backend, like back-sql(5), may fully support them; others may only support a portion of the described semantics, or even differ in some aspects. The relevant details are described in the backend-specific man pages.


It is strongly recommended to explicitly use the most appropriate <dnstyle> in <what> and <who> clauses, to avoid possible incorrect specifications of the access rules as well as for performance (avoid unnecessary regex matching when an exact match suffices) reasons.

An administrator might create a rule of the form:

         access to dn.regex="dc=example,dc=com"
                 by ...

expecting it to match all entries in the subtree "dc=example,dc=com". However, this rule actually matches any DN which contains anywhere the substring "dc=example,dc=com". That is, the rule matches both "uid=joe,dc=example,dc=com" and "dc=example,dc=com,uid=joe".

To match the desired subtree, the rule would be more precisely written:

         access to dn.regex="^(.+,)?dc=example,dc=com$"
                 by ...

For performance reasons, it would be better to use the subtree style.

         access to dn.subtree="dc=example,dc=com"
                 by ...

When writing submatch rules, it may be convenient to avoid unnecessary regex <dnstyle> use; for instance, to allow access to the subtree of the user that matches the <what> clause, one could use

         access to dn.regex="^(.+,)?uid=([^,]+),dc=example,dc=com$"
                 by dn.regex="^uid=$2,dc=example,dc=com$$" write
                 by ...

However, since all that is required in the <by> clause is substring expansion, a more efficient solution is

         access to dn.regex="^(.+,)?uid=([^,]+),dc=example,dc=com$"
                 by dn.exact,expand="uid=$2,dc=example,dc=com" write
                 by ...

In fact, while a <dnstyle> of regex implies substring expansion, exact, as well as all the other DN specific <dnstyle> values, does not, so it must be explicitly requested.


default slapd configuration file


slapd(8), slapd-*(5), slapacl(8), regex(7), re_format(7)

"OpenLDAP Administrator's Guide" (


OpenLDAP Software is developed and maintained by The OpenLDAP Project <>. OpenLDAP Software is derived from University of Michigan LDAP 3.3 Release.