.lf 1 stdin
.TH SLAPD.ACCESS 5 "2020/04/28" "OpenLDAP 2.4.50"
.\" Copyright 1998-2020 The OpenLDAP Foundation All Rights Reserved.
.\" Copying restrictions apply.  See COPYRIGHT/LICENSE.
.\" $OpenLDAP$
.SH NAME
slapd.access \- access configuration for slapd, the stand-alone LDAP daemon
.SH SYNOPSIS
/etc/openldap/slapd.conf
.SH DESCRIPTION
The 
.BR slapd.conf (5)
file contains configuration information for the
.BR slapd (8)
daemon. This configuration file is also used by the SLAPD tools
.BR slapacl (8),
.BR slapadd (8),
.BR slapauth (8),
.BR slapcat (8),
.BR slapdn (8),
.BR slapindex (8),
and
.BR slaptest (8).
.LP
The
.B slapd.conf
file consists of a series of global configuration options that apply to
.B slapd
as a whole (including all backends), followed by zero or more database
backend definitions that contain information specific to a backend
instance.
.LP
The general format of
.B slapd.conf
is as follows:
.LP
.nf
    # 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
    ...
.fi
.LP
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.
.LP
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").
.LP
When dealing with an access list, because the global access list is 
effectively appended to each per-database list, if the resulting 
list is non-empty then the access list will end with an implicit 
.B access to * by * none
directive. If there are no access directives applicable to a backend, 
then a default read is used.
.LP
.B Be warned: the rootdn can always read and write EVERYTHING!
.LP
For entries not held in any backend (such as a root DSE), the
global directives are used.
.LP
Arguments that should be replaced by actual text are shown in
brackets <>.
.SH THE ACCESS DIRECTIVE
The structure of the access control directives is
.TP
.B access to <what> "[ by <who> [ <access> ] [ <control> ] ]+"
Grant access (specified by 
.BR <access> ) 
to a set of entries and/or attributes (specified by 
.BR <what> ) 
by one or more requestors (specified by 
.BR <who> ).

.LP
Lists of access directives are evaluated in the order they appear
in \fIslapd.conf\fP.
When a
.B <what>
clause matches the datum whose access is being evaluated, its
.B <who>
clause list is checked.
When a
.B <who>
clause matches the accessor's properties, its
.B <access>
and
.B <control>
clauses are evaluated.
Access control checking stops at the first match of the
.B <what>
and
.B <who>
clause, unless otherwise dictated by the
.B <control>
clause.
Each
.B <who>
clause list is implicitly terminated by a
.LP
.nf
	by * none stop
.fi
.LP
clause that results in stopping the access control with no access 
privileges granted.
Each
.B <what>
clause list is implicitly terminated by a
.LP
.nf
	access to *
		by * none
.fi
.LP
clause that results in granting no access privileges to an otherwise
unspecified datum.
.SH THE <WHAT> FIELD
The field
.BR <what>
specifies the entity the access control directive applies to.
It can have the forms
.LP
.nf
	dn[.<dnstyle>]=<dnpattern>
	filter=<ldapfilter>
	attrs=<attrlist>[ val[/matchingRule][.<attrstyle>]=<attrval>]
.fi
.LP
with
.LP
.nf
	<dnstyle>={{exact|base(object)}|regex
		|one(level)|sub(tree)|children}
	<attrlist>={<attr>|[{!|@}]<objectClass>}[,<attrlist>]
	<attrstyle>={{exact|base(object)}|regex
		|one(level)|sub(tree)|children}
.fi
.LP
The statement
.B dn=<dnpattern>
selects the entries based on their naming context.
The
.B <dnpattern>
is a string representation of the entry's DN.
The wildcard
.B *
stands for all the entries, and it is implied if no
.B dn
form is given.
.LP
The 
.B <dnstyle> 
is optional; however, it is recommended to specify it to avoid ambiguities.
.B Base 
(synonym of
.BR baseObject ),
the default,
or
.B exact 
(an alias of 
.BR base )
indicates the entry whose DN is equal to the
.BR <dnpattern> ;
.B one
(synonym of
.BR onelevel )
indicates all the entries immediately below the
.BR <dnpattern> ,
.B sub
(synonym of
.BR subtree )
indicates all entries in the subtree at the
.BR <dnpattern> ,
.B children
indicates all the entries below (subordinate to) the 
.BR <dnpattern> .
.LP
If the
.B <dnstyle>
qualifier is
.BR regex ,
then 
.B <dnpattern>
is a POSIX (''extended'') regular expression pattern,
as detailed in
.BR regex (7)
and/or
.BR 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.
.LP
The statement
.B filter=<ldapfilter>
selects the entries based on a valid LDAP filter as described in RFC 4515.
A filter of
.B (objectClass=*)
is implied if no
.B filter
form is given.
.LP
The statement
.B attrs=<attrlist>
selects the attributes the access control rule applies to.
It is a comma-separated list of attribute types, plus the special names
.BR entry ,
indicating access to the entry itself, and
.BR 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 
.B <attrlist>
that are prefixed by
.B @
are directly treated as objectClass names.  A name prefixed by
.B !
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
.B attrs
form is given, 
.B attrs=@extensibleObject
is implied, i.e. all attributes are addressed.
.LP
Using the form
.B 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
.B <attrstyle>
.B 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
.B <attrstyle>
is
.BR regex ,
the provided value is used as a POSIX (''extended'') regular
expression pattern.  If the attribute has DN syntax, the 
.B <attrstyle>
can be any of
.BR base ,
.BR onelevel ,
.B subtree
or
.BR children ,
resulting in base, onelevel, subtree or children match, respectively.
.LP
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.
Submatches resulting from
.B regex
matching can be dereferenced in the
.B <who>
field using the syntax
.IR ${v<n>} ,
where
.I <n>
is the submatch number.
The default syntax,
.IR $<n> ,
is actually an alias for
.IR ${d<n>} ,
that corresponds to dereferencing submatches from the
.B dnpattern
portion of the
.B <what>
field.
.SH THE <WHO> FIELD
The field
.B <who>
indicates whom the access rules apply to.
Multiple 
.B <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
.LP
.nf
	*
	anonymous
	users
	self[.<selfstyle>]

	dn[.<dnstyle>[,<modifier>]]=<DN>
	dnattr=<attrname>

	realanonymous
	realusers
	realself[.<selfstyle>]

	realdn[.<dnstyle>[,<modifier>]]=<DN>
	realdnattr=<attrname>

	group[/<objectclass>[/<attrname>]]
		[.<groupstyle>]=<group>
	peername[.<peernamestyle>]=<peername>
	sockname[.<style>]=<sockname>
	domain[.<domainstyle>[,<modifier>]]=<domain>
	sockurl[.<style>]=<sockurl>
	set[.<setstyle>]=<pattern>

	ssf=<n>
	transport_ssf=<n>
	tls_ssf=<n>
	sasl_ssf=<n>

	dynacl/<name>[/<options>][.<dynstyle>][=<pattern>]
.fi
.LP
with
.LP
.nf
	<style>={exact|regex|expand}
	<selfstyle>={level{<n>}}
	<dnstyle>={{exact|base(object)}|regex
		|one(level)|sub(tree)|children|level{<n>}}
	<groupstyle>={exact|expand}
	<peernamestyle>={<style>|ip|ipv6|path}
	<domainstyle>={exact|regex|sub(tree)}
	<setstyle>={exact|expand}
	<modifier>={expand}
	<name>=aci		<pattern>=<attrname>]
.fi
.LP
They may be specified in combination.
.LP
.nf
.fi
.LP
The wildcard
.B *
refers to everybody.
.LP
The keywords prefixed by
.B real
act as their counterparts without prefix; the checking respectively occurs
with the \fIauthentication\fP DN and the \fIauthorization\fP DN.
.LP
The keyword
.B anonymous
means access is granted to unauthenticated clients; it is mostly used 
to limit access to authentication resources (e.g. the
.B userPassword
attribute) to unauthenticated clients for authentication purposes.
.LP
The keyword
.B users
means access is granted to authenticated clients.
.LP
The keyword
.B 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 
.B level{<n>}
style, where \fI<n>\fP 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 "\fIby self.level{1} ...\fP" clause would match
when the object "\fIdc=example,dc=com\fP" is accessed
by "\fIcn=User,dc=example,dc=com\fP".
A "\fIby self.level{-1} ...\fP" clause would match when the same user
accesses the object "\fIou=Address Book,cn=User,dc=example,dc=com\fP".
.LP
The statement
.B dn=<DN>
means that access is granted to the matching DN.
The optional style qualifier
.B dnstyle
allows the same choices of the dn form of the
.B <what>
field.  In addition, the
.B regex
style can exploit substring substitution of submatches in the
.B <what>
dn.regex clause by using the form
.BR $<digit> ,
with 
.B digit
ranging from 0 to 9 (where 0 matches the entire string),
or the form
.BR ${<digit>+} ,
for submatches higher than 9.
Substring substitution from attribute value can
be done in 
using the form
.BR ${v<digit>+} .
Since the dollar character is used to indicate a substring replacement,
the dollar character that is used to indicate match up to the end of
the string must be escaped by a second dollar character, e.g.
.LP
.nf
    access to dn.regex="^(.+,)?uid=([^,]+),dc=[^,]+,dc=com$"
        by dn.regex="^uid=$2,dc=[^,]+,dc=com$$" write
.fi
.LP
The style qualifier
allows an optional
.BR modifier .
At present, the only type allowed is 
.BR expand ,
which causes substring substitution of submatches to take place
even if 
.B dnstyle
is not 
.BR regex .
Note that the 
.B regex 
dnstyle in the above example may be of use only if the 
.B <by>
clause needs to be a regex; otherwise, if the
value of the second (from the right)
.B dc=
portion of the DN in the above example were fixed, the form
.LP
.nf
    access to dn.regex="^(.+,)?uid=([^,]+),dc=example,dc=com$"
        by dn.exact,expand="uid=$2,dc=example,dc=com" write
.fi
.LP
could be used; if it had to match the value in the 
.B <what>
clause, the form
.LP
.nf
    access to dn.regex="^(.+,)?uid=([^,]+),dc=([^,]+),dc=com$"
        by dn.exact,expand="uid=$2,dc=$3,dc=com" write
.fi
.LP
could be used.
.LP
Forms of the 
.B <what>
clause other than regex may provide submatches as well.
The 
.BR base(object) ,
the
.BR sub(tree) ,
the
.BR one(level) ,
and the
.BR children
forms provide
.B $0
as the match of the entire string.
The 
.BR sub(tree) ,
the
.BR one(level) ,
and the
.BR children
forms also provide
.B $1
as the match of the rightmost part of the DN as defined in the
.B <what>
clause.
This may be useful, for instance, to provide access to all the 
ancestors of a user by defining
.LP
.nf
    access to dn.subtree="dc=com"
        by dn.subtree,expand="$1" read
.fi
.LP
which means that only access to entries that appear in the DN of the
.B <by>
clause is allowed.
.LP
The 
.BR level{<n>}
form is an extension and a generalization of the
.BR onelevel
form, which matches all DNs whose <n>-th ancestor is the pattern.
So, \fIlevel{1}\fP is equivalent to \fIonelevel\fP, 
and \fIlevel{0}\fP is equivalent to \fIbase\fP.
.LP
It is perfectly useless to give any access privileges to a DN 
that exactly matches the
.B 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
.BR rootdn ,
to solve the intrinsic chicken-and-egg problem.
.LP
The statement
.B dnattr=<attrname>
means that access is granted to requests whose DN is listed in the
entry being accessed under the 
.B <attrname>
attribute.
.LP
The statement
.B group=<group>
means that access is granted to requests whose DN is listed
in the group entry whose DN is given by
.BR <group> .
The optional parameters
.B <objectclass>
and
.B <attrname>
define the objectClass and the member attributeType of the group entry.
The defaults are
.B groupOfNames
and
.BR member ,
respectively.
The optional style qualifier
.B <style>
can be
.BR expand ,
which means that
.B <group>
will be expanded as a replacement string (but not as a regular expression)
according to
.BR regex (7)
and/or
.BR re_format (7),
and
.BR exact ,
which means that exact match will be used.
If the style of the DN portion of the
.B <what>
clause is regex, the submatches are made available according to
.BR regex (7)
and/or
.BR re_format (7);
other styles provide limited submatches as discussed above about 
the DN form of the 
.B <by>
clause.
.LP
For static groups, the specified attributeType must have
.B DistinguishedName
or
.B NameAndOptionalUID
syntax. For dynamic groups the attributeType must
be a subtype of the
.B labeledURI
attributeType. Only LDAP URIs of the form
.B ldap:///<base>??<scope>?<filter>
will be evaluated in a dynamic group, by searching the local server only.
.LP
The statements
.BR peername=<peername> ,
.BR sockname=<sockname> ,
.BR domain=<domain> ,
and
.BR sockurl=<sockurl>
mean that the contacting host IP (in the form 
.BR "IP=<ip>:<port>"
for IPv4, or
.BR "IP=[<ipv6>]:<port>"
for IPv6)
or the contacting host named pipe file name (in the form
.B "PATH=<path>"
if connecting through a named pipe) for
.BR peername ,
the named pipe file name for
.BR sockname ,
the contacting host name for
.BR domain ,
and the contacting URL for
.BR sockurl
are compared against
.B pattern
to determine access.
The same
.B style
rules for pattern match described for the
.B group
case apply, plus the
.B regex
style, which implies submatch
.B expand
and regex match of the corresponding connection parameters.
The
.B exact
style of the
.BR <peername>
clause (the default) implies a case-exact match on the client's
.BR IP , 
including the
.B "IP="
prefix and the trailing
.BR ":<port>" , 
or the client's 
.BR path ,
including the
.B "PATH="
prefix if connecting through a named pipe.
The special
.B ip
style interprets the pattern as 
.BR <peername>=<ip>[%<mask>][{<n>}] ,
where
.B <ip>
and 
.B <mask>
are dotted digit representations of the IP and the mask, while
.BR <n> ,
delimited by curly brackets, is an optional port.
The same applies to IPv6 addresses when the special
.B ipv6
style is used.
When checking access privileges, the IP portion of the
.BR peername 
is extracted, eliminating the
.B "IP="
prefix and the
.B ":<port>"
part, and it is compared against the
.B <ip>
portion of the pattern after masking with
.BR <mask> :
\fI((peername & <mask>) == <ip>)\fP.
As an example, 
.B peername.ip=127.0.0.1
and
.B peername.ipv6=::1
allow connections only from localhost,
.B peername.ip=192.168.1.0%255.255.255.0 
allows connections from any IP in the 192.168.1 class C domain, and
.B peername.ip=192.168.1.16%255.255.255.240{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 
.B path
style eliminates the 
.B "PATH="
prefix from the 
.B peername
when connecting through a named pipe, and performs an exact match 
on the given pattern.
The
.BR <domain>
clause also allows the
.B subtree
style, which succeeds when a fully qualified name exactly matches the
.BR domain
pattern, or its trailing part, after a 
.BR dot ,
exactly matches the 
.BR domain
pattern.
The 
.B expand
style is allowed, implying an
.B exact 
match with submatch expansion; the use of 
.B expand 
as a style modifier is considered more appropriate.
As an example,
.B domain.subtree=example.com
will match www.example.com, but will not match www.anotherexample.com.
The
.B domain
of the contacting host is determined by performing a DNS reverse lookup.
As this lookup can easily be spoofed, use of the
.B domain
statement is strongly discouraged.  By default, reverse lookups are disabled.
The optional
.B domainstyle
qualifier of the
.B <domain>
clause allows a
.B modifier
option; the only value currently supported is
.BR expand ,
which causes substring substitution of submatches to take place even if
the 
.B domainstyle
is not 
.BR regex ,
much like the analogous usage in 
.B <dn>
clause.
.LP
The statement
.B set=<pattern>
is undocumented yet.
.LP
The statement
.B dynacl/<name>[/<options>][.<dynstyle>][=<pattern>]
means that access checking is delegated to the admin-defined method
indicated by
.BR <name> ,
which can be registered at run-time by means of the
.B moduleload
statement.
The fields
.BR <options> ,
.B <dynstyle>
and
.B <pattern>
are optional, and are directly passed to the registered parsing routine.
Dynacl is experimental; it must be enabled at compile time.
.LP
The statement
.B dynacl/aci[=<attrname>]
means that the access control is determined by the values in the
.B attrname
of the entry itself.
The optional
.B <attrname>
indicates what attributeType holds the ACI information in the entry.
By default, the 
.B OpenLDAPaci
operational attribute is used.
ACIs are experimental; they must be enabled at compile time.
.LP
The statements
.BR ssf=<n> ,
.BR transport_ssf=<n> ,
.BR tls_ssf=<n> ,
and
.BR sasl_ssf=<n>
set the minimum required Security Strength Factor (ssf) needed
to grant access.  The value should be positive integer.
.SH THE <ACCESS> FIELD
The optional field
.B <access> ::= [[real]self]{<level>|<priv>}
determines the access level or the specific access privileges the
.B who 
field will have.
Its component are defined as
.LP
.nf
	<level> ::= none|disclose|auth|compare|search|read|{write|add|delete}|manage
	<priv> ::= {=|+|\-}{0|d|x|c|s|r|{w|a|z}|m}+
.fi
.LP
The modifier
.B 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
.B realself
refers to the authenticated DN as opposed to the authorized DN of the
.B self
modifier.
An example is the
.B 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, while being not allowed
to affect other members.
.LP
The 
.B level 
access model relies on an incremental interpretation of the access
privileges.
The possible levels are
.BR none ,
.BR disclose ,
.BR auth ,
.BR compare ,
.BR search ,
.BR read ,
.BR write ,
and
.BR manage .
Each access level implies all the preceding ones, thus 
.B manage
grants all access including administrative access.
The 
.BR write
access is actually the combination of
.BR add
and
.BR delete ,
which respectively restrict the write privilege to add or delete
the specified
.BR <what> .

.LP
The
.B none 
access level disallows all access including disclosure on error.
.LP
The
.B disclose
access level allows disclosure of information on error.
.LP
The 
.B auth
access level means that one is allowed access to an attribute to perform
authentication/authorization operations (e.g.
.BR bind )
with no other access.
This is useful to grant unauthenticated clients the least possible 
access level to critical resources, like passwords.
.LP
The
.B priv
access model relies on the explicit setting of access privileges
for each clause.
The
.B =
sign resets previously defined accesses; as a consequence, the final 
access privileges will be only those defined by the clause.
The 
.B +
and
.B \-
signs add/remove access privileges to the existing ones.
The privileges are
.B m
for manage,
.B w
for write,
.B a
for add,
.B z
for delete,
.B r
for read,
.B s 
for search,
.B c 
for compare,
.B x
for authentication, and
.B d
for disclose.
More than one of the above privileges can be added in one statement.
.B 0
indicates no privileges and is used only by itself (e.g., +0).
Note that
.B +az
is equivalent to
.BR +w .
.LP
If no access is given, it defaults to 
.BR +0 .
.SH THE <CONTROL> FIELD
The optional field
.B <control>
controls the flow of access rule application.
It can have the forms
.LP
.nf
	stop
	continue
	break
.fi
.LP
where
.BR 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
.B continue
form allows for other 
.B <who>
clauses in the same 
.B <access>
clause to be considered, so that they may result in incrementally altering
the privileges, while the
.B break
form allows for other
.B <access>
clauses that match the same target to be processed.
Consider the (silly) example
.LP
.nf
	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
.fi
.LP
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
.LP
.nf
	access to dn.subtree="dc=example,dc=com" attrs=cn
		by * =cs continue
		by users +r
.fi
.LP
which grants everybody search and compare privileges, and adds read
privileges to authenticated clients.
.LP
One useful application is to easily grant write privileges to an
.B updatedn
that is different from the
.BR rootdn .
In this case, since the
.B updatedn
needs write access to (almost) all data, one can use
.LP
.nf
	access to *
		by dn.exact="cn=The Update DN,dc=example,dc=com" write
		by * break
.fi
.LP
as the first access rule.
As a consequence, unless the operation is performed with the 
.B updatedn
identity, control is passed straight to the subsequent rules.

.SH OPERATION REQUIREMENTS
Operations require different privileges on different portions of entries.
The following summary applies to primary MDB database backend and the
deprecated BDB and HDB backends.   Requirements for other backends may
(and often do) differ.

.LP
The
.B add
operation requires
.B add (=a)
privileges on the pseudo-attribute 
.B entry
of the entry being added, and 
.B add (=a)
privileges on the pseudo-attribute
.B children
of the entry's parent.
When adding the suffix entry of a database,
.B add
access to
.B children
of the empty DN ("") is required. Also if
Add content ACL checking has been configured on
the database (see the
.BR slapd.conf (5)
or
.BR slapd\-config (5)
manual page),
.B add (=a)
will be required on all of the attributes being added.

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

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

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

.LP
The
.B modify
operation requires 
.B write (=w)
privileges on the attributes being modified.
In detail, 
.B add (=a)
is required to add new values,
.B delete (=z)
is required to delete existing values,
and both
.B delete
and
.BR "add (=az)" ,
or
.BR "write (=w)" ,
are required to replace existing values.

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

.LP
The
.B search
operation, requires 
.B search (=s)
privileges on the 
.B entry
pseudo-attribute of the searchBase
(NOTE: this was introduced with OpenLDAP 2.4).
Then, for each entry, it requires
.B search (=s)
privileges on the attributes that are defined in the filter.
The resulting entries are finally tested for 
.B read (=r)
privileges on the pseudo-attribute
.B entry
(for read access to the entry itself)
and for
.B read (=r)
access on each value of each attribute that is requested.
Also, for each
.B referral
object used in generating continuation references, the operation requires
.B read (=r)
access on the pseudo-attribute
.B entry
(for read access to the referral object itself),
as well as
.B read (=r)
access to the attribute holding the referral information
(generally the
.B ref
attribute).

.LP
Some internal operations and some
.B controls
require specific access privileges.
The
.B authzID
mapping and the 
.B proxyAuthz
control require
.B 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
.B authz-regexp
directives).
.B Auth (=x)
privileges are also required on the
.B authzTo
attribute of the authorizing identity and/or on the 
.B authzFrom
attribute of the authorized identity.
In general, when an internal lookup is performed for authentication
or authorization purposes, search-specific privileges (see the access
requirements for the search operation illustrated above) are relaxed to
.BR auth .

.LP
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.
.BR back\-bdb (5),
and
.BR back\-hdb (5).

Some other backend, like
.BR 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.

.SH CAVEATS
It is strongly recommended to explicitly use the most appropriate
.B <dnstyle>
in
.B <what>
and
.B <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.
.LP
An administrator might create a rule of the form:
.LP
.nf
	access to dn.regex="dc=example,dc=com"
		by ...
.fi
.LP
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".
.LP
To match the desired subtree, the rule would be more precisely
written:
.LP
.nf
	access to dn.regex="^(.+,)?dc=example,dc=com$"
		by ...
.fi
.LP
For performance reasons, it would be better to use the subtree style.
.LP
.nf
	access to dn.subtree="dc=example,dc=com"
		by ...
.fi
.LP
When writing submatch rules, it may be convenient to avoid unnecessary
.B regex
.B <dnstyle>
use; for instance, to allow access to the subtree of the user 
that matches the
.B <what>
clause, one could use
.LP
.nf
	access to dn.regex="^(.+,)?uid=([^,]+),dc=example,dc=com$"
		by dn.regex="^uid=$2,dc=example,dc=com$$" write
		by ...
.fi
.LP
However, since all that is required in the 
.B <by>
clause is substring expansion, a more efficient solution is
.LP
.nf
	access to dn.regex="^(.+,)?uid=([^,]+),dc=example,dc=com$"
		by dn.exact,expand="uid=$2,dc=example,dc=com" write
		by ...
.fi
.LP
In fact, while a
.B <dnstyle>
of
.B regex
implies substring expansion, 
.BR exact ,
as well as all the other DN specific
.B <dnstyle>
values, does not, so it must be explicitly requested.
.LP
.SH FILES
.TP
/etc/openldap/slapd.conf
default slapd configuration file
.SH SEE ALSO
.BR slapd (8),
.BR slapd\-* (5),
.BR slapacl (8),
.BR regex (7),
.BR re_format (7)
.LP
"OpenLDAP Administrator's Guide" (http://www.OpenLDAP.org/doc/admin/)
.SH ACKNOWLEDGEMENTS
.lf 1 ./../Project
.\" Shared Project Acknowledgement Text
.B "OpenLDAP Software"
is developed and maintained by The OpenLDAP Project <http://www.openldap.org/>.
.B "OpenLDAP Software"
is derived from the University of Michigan LDAP 3.3 Release.  
.lf 1184 stdin