KRB5.CONF(5)

📦
version 7.8.git20240113.1b4565a · 3 versions
, in debian trixie, ubuntu questing, ubuntu resolute, ubuntu stonking

✨ NAME ✨

okay so bestie, krb5.conf is literally the configuration file for Kerberos 5. 💖 And this manual page? It's KRB5.CONF(5) — we're just folding that name in right here, no cap. ✨

🌸 SYNOPSIS 🌸

#include <krb5.h>

💖 DESCRIPTION 💖

The krb5.conf file is where you specify a bunch of configuration parameters for the Kerberos 5 library, and also for some programs. It's giving main character energy for your Kerberos setup, fr fr. 🎀

The file is made of one or more sections, and each section has a bunch of bindings. The value of each binding can be either a string or a list of other bindings. The grammar looks like this:

             file:
                     /* empty */
                     sections
                     includes

             sections:
                     section sections
                     section

             section:
                     '[' section_name ']' bindings

             section_name:
                     STRING

             bindings:
                     binding bindings
                     binding

             binding:
                     name '=' STRING
                     name '=' '{' bindings '}'

             name:
                     STRING

             includes:
                     'include' path
                     'includedir' path

             path: STRING

A STRING is just one or more non-whitespace characters. Simple as that. 💅

Files and directories can be included by absolute path. If you include a directory, all files in that directory get included as if each file had been included separately. But only files whose names consist of alphanumeric, hyphen, and underscore are included, though they may also end in .conf. So like, don't name your files with spaces or emoji, bestie. 🙈

STRINGs that are specified later in this man-page use the following notation:

  • boolean — values can be either yes/true or no/false.
  • time — values can be a list of year, month, day, hour, min, second. Example: 1 month 2 days 30 min. If no unit is given, seconds is assumed.
  • etypes — valid encryption types are: des-cbc-crc, des-cbc-md4, des-cbc-md5, des3-cbc-sha1, arcfour-hmac-md5, aes128-cts-hmac-sha1-96, and aes256-cts-hmac-sha1-96 .
  • address — an address can be either a IPv4 or a IPv6 address.

Currently recognised sections and bindings are:

[appdefaults]

Specifies the default values to be used for Kerberos applications. You can specify defaults per application, realm, or a combination of these. The preference order is:

  1. application realm option
  2. application option
  3. realm option
  4. option

The supported options are:

  • forwardable = boolean — When obtaining initial credentials, make the credentials forwardable. (slay, now they can travel 💅)
  • proxiable = boolean — When obtaining initial credentials, make the credentials proxiable.
  • no-addresses = boolean — When obtaining initial credentials, request them for an empty set of addresses, making the tickets valid from any address.
  • ticket_lifetime = time — Default ticket lifetime.
  • renew_lifetime = time — Default renewable ticket lifetime.
  • encrypt = boolean — Use encryption, when available.
  • forward = boolean — Forward credentials to remote host (for rsh(1), telnet(1), etc).
  • historical_anon_pkinit = boolean — Enable legacy anonymous pkinit command-line syntax. With this option set to true, the kinit(1) --anonymous command with no principal argument specified will request an anonymous pkinit ticket from the default realm. If a principal argument is specified, it is used as an explicit realm name for anonymous pkinit even without an @ prefix.

[libdefaults]

  • default_realm = REALM — Default realm to use, this is also known as your “local realm”. The default is the result of krb5_get_host_realm(local hostname).
  • allow_weak_crypto = boolean — are weak crypto algorithms allowed to be used, among others, DES is considered weak.
  • clockskew = time — Maximum time differential (in seconds) allowed when comparing times. Default is 300 seconds (five minutes).
  • kdc_timeout = time — Maximum time to wait for a reply from the kdc, default is 3 seconds.
  • capath — This is deprecated, see the capaths section below. (not me using deprecated stuff 🙈)
capath = {

           destination-realm = next-hop-realm

           ...

           }
  • default_cc_type = cctype — sets the default credentials type.
  • default_cc_name = ccname — the default credentials cache name. If you want to change the type only use default_cc_type. The string can contain variables that are expanded on runtime. The only supported variable currently is %{uid} which expands to the current user id.
  • default_etypes = etypes ... — A list of default encryption types to use. (Default: all enctypes if allow_weak_crypto = TRUE, else all enctypes except single DES enctypes.)
  • default_as_etypes = etypes ... — A list of default encryption types to use in AS requests. (Default: the value of default_etypes.)
  • default_tgs_etypes = etypes ... — A list of default encryption types to use in TGS requests. (Default: the value of default_etypes.)
  • default_etypes_des = etypes ... — A list of default encryption types to use when requesting a DES credential.
  • default_keytab_name = keytab — The keytab to use if no other is specified, default is “FILE:/etc/krb5.keytab”.
  • dns_lookup_kdc = boolean — Use DNS SRV records to lookup KDC services location.
  • dns_lookup_realm = boolean — Use DNS TXT records to lookup domain to realm mappings.
  • kdc_timesync = boolean — Try to keep track of the time differential between the local machine and the KDC, and then compensate for that when issuing requests.
  • max_retries = number — The max number of times to try to contact each KDC.
  • large_msg_size = number — The threshold where protocols with tiny maximum message sizes are not considered usable to send messages to the KDC.
  • ticket_lifetime = time — Default ticket lifetime.
  • renew_lifetime = time — Default renewable ticket lifetime.
  • forwardable = boolean — When obtaining initial credentials, make the credentials forwardable. This option is also valid in the [realms] section.
  • proxiable = boolean — When obtaining initial credentials, make the credentials proxiable. This option is also valid in the [realms] section.
  • verify_ap_req_nofail = boolean — If enabled, failure to verify credentials against a local key is a fatal error. The application has to be able to read the corresponding service key for this to work. Some applications, like su(1), enable this option unconditionally.
  • warn_pwexpire = time — How soon to warn for expiring password. Default is seven days.
  • http_proxy = proxy-spec — A HTTP-proxy to use when talking to the KDC via HTTP.
  • dns_proxy = proxy-spec — Enable using DNS via HTTP.
  • extra_addresses = address ... — A list of addresses to get tickets for along with all local addresses.
  • time_format = string — How to print time strings in logs, this string is passed to strftime(3).
  • date_format = string — How to print date strings in logs, this string is passed to strftime(3).
  • log_utc = boolean — Write log-entries using UTC instead of your local time zone.
  • scan_interfaces = boolean — Scan all network interfaces for addresses, as opposed to simply using the address associated with the system's host name.
  • fcache_version = int — Use file credential cache format version specified.
  • fcc-mit-ticketflags = boolean — Use MIT compatible format for file credential cache. It's the field ticketflags that is stored in reverse bit order for older than Heimdal 0.7. Setting this flag to TRUE makes it store the MIT way, this is default for Heimdal 0.7.
  • check-rd-req-server — If set to "ignore", the framework will ignore any of the server input to krb5_rd_req(3), this is very useful when the GSS-API server input the wrong server name into the gss_accept_sec_context call.
  • k5login_directory = directory — a whole alternate location for user .k5login files. It's here purely for MIT krb5 config-file compatibility, bestie, so if you're migrating, your old file still vibes. ✨
  • k5login_authoritative = boolean — if this is true, then when a principal isn't found in k5login files, krb5_userok(3) will not fall back on principal-to-username mapping. Also just MIT krb5 compatibility. 🫶
  • kuserok = rule ... — this is the main character setting the whole krb5_userok(3) behaviour. Give it multiple values and krb5_userok(3) evaluates them in order until one succeeds, or all of them fail. 😭 Rules come from plugins, and three built-in plugins are described right below. Default: USER-K5LOGIN SIMPLE DENY.
    • kuserok = DENY — if this gets evaluated, krb5_userok(3) denies access to the given username no matter what the principal name is. Periodt, no exceptions. 💀
    • kuserok = SIMPLE — if this gets evaluated, krb5_userok(3) uses principal-to-username mapping (peep auth_to_local below 👀). If the principal maps to the requested username, access is allowed. Slay.
    • kuserok = SYSTEM-K5LOGIN[:directory] — if this gets evaluated, krb5_userok(3) uses k5login files named after the luser argument to krb5_userok(3), living in the given directory or in /etc/k5login.d/. K5login files are text files, one principal name per line, and principals appearing in a user's k5login file are permitted access to that user's account. 👑
Heads up bestie: this rule performs no ownership checks and no permissions checks on k5login files. Proper ownership and permissions/ACLs are expected because the k5login location is a system location. It's giving "trust me bro" and that's on you. 🙈
  • kuserok = USER-K5LOGIN — if this gets evaluated, krb5_userok(3) uses ~luser/.k5login and ~luser/.k5login.d/*. User k5login files and directories must be owned by the user and must not have world or group write permissions. 🌸
  • aname2lname-text-db = filename — the named file has to be a sorted (increasing order, no chaos) text file where every line is an unparsed principal name, optionally followed by whitespace and a username. The aname2lname function binary-searches this file when it's configured, looking for lines matching the given principal name. Found it? Then the given username gets used — or, if the username is missing, an error comes back. If the file doesn't exist, or no matching line turns up, then other plugins get to run. 💅
  • fcache_strict_checking — strict checking in FILE credential caches, so that owner, no-symlink and permissions are all correct. Vibe check for your ccache, basically. 🔥
  • name_canon_rules = rules — one or more service principal name canonicalization rules. Each rule is one or more tokens separated by colon (':'). Currently these rules are used only for hostname canonicalization — usually when getting a service ticket, from a ccache or a TGS, but also when acquiring GSS initiator credentials from a keytab. You can use them to implement DNS-resolver-style search lists without actually having to use DNS. ✨
NOTE: Name canonicalization rules are an experimental feature. 🧪

The first token is a rule type, one of: as-is, qualify, or nss.

Any remaining tokens must be options tokens: use_fast (use FAST to protect TGS exchanges; currently not supported), use_dnssec (use DNSSEC to protect hostname lookups; currently not supported), ccache_only , use_referrals, no_referrals, lookup_realm, mindots=N, maxdots=N, order=N, domain= domain, realm= realm, match_domain= domain, and match_realm= realm.

When trying to obtain a service ticket for a host-based service principal name, name canonicalization rules get applied to that name in the order given, one by one, until one succeeds (a service ticket is obtained), or all of them fail. Same deal when acquiring GSS initiator credentials from a keytab, and when comparing a non-canonical GSS name to a canonical one. 🦋

For each rule the system checks that the hostname has at least mindots periods (if given) in it, at most maxdots periods (if given), that the hostname ends in the given match_domain (if given), and that the realm of the principal matches the match_realm (if given).

  • as-is rules leave the hostname unmodified but may set a realm.
  • qualify rules qualify the hostname with the given domain and also may set the realm.
  • The nss rule uses the system resolver to look up the host's canonical name and is usually not secure. Note that using the nss rule type implies having to have principal aliases in the HDB (though not necessarily in keytabs). ⚠️

The empty realm denotes "ask the client's realm's TGS". The empty realm may be set as well as matched. 💖

The order in which rules are applied goes like this: first all the rules with explicit order, then all other rules in the order in which they appear. If any two rules have the same explicit order, their order of appearance in krb5.conf breaks the tie. Explicitly specifying order can be useful where tools read and write the configuration file without preserving parameter order. (Not me side-eyeing tools that scramble my config 🙄.)

Malformed rules are ignored.

  • allow_hierarchical_capaths = boolean — when validating cross-realm transit paths, absent any explicit capath from the client realm to the server realm, this allows a hierarchical transit path via the common ancestor domain of the two realms. Defaults to true. Note: absent an explicit setting, hierarchical capaths are always used by the KDC when generating a referral to a destination with which there is no direct trust. 🌈

🌸 [domain_realm] 🌸

This one is a list of mappings from DNS domain to Kerberos realm. Each binding in this section looks like:

domain = realm

The domain can be either a full name of a host or a trailing component. In the trailing-component case, the domain-string should start with a period. The trailing component only matches hosts that are in the same domain, ie .example.com matches foo.example.com, but not foo.test.example.com. 🎯

The realm may be the token dns_locate, in which case the actual realm will be determined using DNS (independently of the setting of the dns_lookup_realm option). 🦋

💜 [realms] 💜

REALM = {
  • kdc = [service/]host[:port] — specifies a list of kdcs for this realm. If the optional port is absent, the default value for the kerberos/udp, kerberos/tcp, and http/tcp port (depending on service) will be used. The kdcs will be used in the order that they are specified. 👑

The optional service specifies over what medium the kdc should be contacted. Possible services are udp, tcp, and http. Http can also be written as http://. Default service is udp and tcp. ✨

  • admin_server = host[:port] — specifies the admin server for this realm, where all the modifications to the database are performed. The main character desk where the edits happen. 💅
  • kpasswd_server = host[:port] — points to the server where all the password changes are performed. If there's no such entry, the kpasswd port on the admin_server host will be tried. 🫶
  • tgs_require_subkey — a boolean variable that defaults to false. Old DCE secd (pre 1.1) might need this to be true. Ancient lore, but okay. 🧚
  • auth_to_local_names = {
  • principal_name = username — you drop an actual principal name here and it gets mapped to the given username, but only if the REALM is a default realm. that's the whole vibe. 💖
  • auth_to_local = HEIMDAL_DEFAULT — use the Heimdal default principal-to-username mapping. Applies to principals from the REALM if and only if REALM is a default realm.
  • auth_to_local = DEFAULT — use the MIT default principal-to-username mapping. Applies to principals from the REALM if and only if REALM is a default realm.
  • auth_to_local = DB:/path/to/db.txt — do a binary search of the given DB. bestie the DB has requirements: it must be a flat-text file sorted in the "C" locale, and each record is a line (separated by either LF or CRLF) made of a principal name, then whitespace, then a username. Applies to principals from the REALM if and only if REALM is a default realm.
  • auth_to_local = DB:/path/to/db — use the given DB, but only if there's a plugin for it. Applies to principals from the REALM if and only if REALM is a default realm.
  • auth_to_local = RULE:... — use the given rule, if there's a plugin for it. Applies to principals from the REALM if and only if REALM is a default realm.
  • auth_to_local = NONE — no extra principal-to-username mapping happens. 👀 heads up though: auth_to_local_names and any preceding auth_to_local rules still take precedence, so this isn't a "delete everything" button.

🦋 [capaths] 🦋

client-realm = {
    server-realm = hop-realm ...
}
  • server-realm = hop-realm ... — okay so this one does two jobs at once, no cap:
    1. The first listed hop-realm tells a client which realm it should hit up in order to eventually get credentials for a service in the server-realm. 🛰️
    2. It also tells the KDC (and other servers) which realms are allowed in a multi-hop traversal from client-realm to server-realm.

And for the second job specifically — the order of the realms doesn't matter. Only the client case cares about ordering. 💅

🌸 [logging] 🌸

  • entity = destination — says that entity should use the given destination for logging. Wanna know all the defined destinations? go peek at the krb5_openlog(3) manual page, it's all listed there. 🎀

👑 [kdc] 👑

database = {
    dbname = [DATBASETYPE:]DATABASENAME
    realm = REALM
    mkey_file = FILENAME
    acl_file = PA FILENAME
    log_file = FILENAME
    log-max-size = number
}
  • dbname = [DATBASETYPE:]DATABASENAME — use this database for this realm. The DATABASETYPE should be one of 'lmdb', 'db3', 'db1', 'db', 'sqlite', or 'ldap'. Wanna know how to actually configure the different database backends? the info documentation has the tea. ☕
  • realm = REALM — says which realm is gonna be stored in this database. If realm isn't set, it gets used as the default database, and there can only be one entry that doesn't have a realm stanza. one. that's it. 🫶
  • mkey_file = FILENAME — use this keytab file for the master key of this database. If you don't specify it, DATABASENAME.mkey gets used instead.
  • acl_file = PA FILENAME — use this file for the ACL list of this database.
  • log_file = FILENAME — use this file as the log of changes made to the database. This file is used by ipropd-master for propagating changes to slaves. It's also used by kadmind and kadmin (when used with the -l option), and by all applications using libkadm5 with the local backend, for two-phase commit functionality. Slaves use it too. Setting this to /dev/null disables two-phase commit and incremental propagation. And iprop-log is your bestie for showing the contents of this log file. ✨
  • log-max-size = number — when the log reaches this size (in bytes), the log gets truncated, saving some entries and keeping the latest version number so incremental propagation isn't disrupted. If it's set to a negative value, automatic log truncation gets disabled. Defaults to 52428800 (50MB).
  • max-request = SIZE — maximum size of a kdc request.
  • require-preauth = BOOL — if set, pre-authentication is required.
  • ports = list of ports — list of ports the kdc should listen to.
  • addresses = list of interfaces — list of addresses the kdc should bind to.
  • enable-http = BOOL — should the kdc answer kdc-requests over http?
  • tgt-use-strongest-session-key = BOOL — if this is TRUE then the KDC will prefer the strongest key from the client's AS-REQ or TGS-REQ enctype list for the ticket session key that is supported by the KDC and the target principal when the target principal is a krbtgt principal. Else it'll prefer the first key from the client's AS-REQ enctype list that's also supported by the KDC and the target principal. Defaults to FALSE.
  • svc-use-strongest-session-key = BOOL — like tgt-use-strongest-session-key, but it applies to the session key enctype of tickets for services other than krbtgt principals. Defaults to FALSE.
  • preauth-use-strongest-session-key = BOOL — if TRUE then it selects the strongest possible enctype from the client's AS-REQ for PA-ETYPE-INFO2 (i.e., for password-based pre-authentication). Else it picks the first supported enctype from the client's AS-REQ. Defaults to FALSE.
  • use-strongest-server-key = BOOL — if TRUE then the KDC picks, for the ticket encrypted part's key, the first supported enctype from the target service principal's hdb entry's current keyset. Else the KDC picks the first supported enctype from the target service principal's hdb entry's current keyset. Defaults to TRUE. 🫠 (yes bestie, both branches really do say the same thing up there — I'm just the messenger, not the author.)
  • check-ticket-addresses = BOOL — verify the addresses in the tickets used in tgs requests.
  • allow-null-ticket-addresses = BOOL — allow address-less tickets.
  • allow-anonymous = BOOL — if the kdc is allowed to hand out anonymous tickets.
  • historical_anon_realm = boolean — enables pre-7.0 non-RFC-comformant KDC behavior. With this option set to true, the client realm in anonymous pkinit AS replies will be the requested realm, rather than the RFC-conformant WELLKNOWN:ANONYMOUS realm. 🚨 This can have a security impact on servers that expect to grant access to anonymous-but-authenticated-to-the-KDC users of the realm in question: they would also grant access to unauthenticated anonymous users. As such, it is not recommended to set this option to true. like, genuinely, don't. 🙈
  • encode_as_rep_as_tgs_rep = BOOL — encode as-rep as tgs-rep to be compatible with mistakes older DCE secd did. 💀
  • kdc_warn_pwexpire = TIME — the time before expiration that the user should be warned that her password is about to expire.
  • logging = Logging — what type of logging the kdc should use. See also [logging]/kdc. 🎀

still in the [kdc] energy bestie, we're not done yet — more LDAP and kx509 lore incoming 💅

  • hdb-ldap-structural-object structural object — if you're using the LDAP backend to stash principals, this is the structural object it grabs when it creates stuff and when it reads stuff back. default value? account . no notes.
  • hdb-ldap-create-base creation dn — this is the dn that gets appended to the principal when it's making entries. default value is the search dn, slay.
  • enable-digest = BOOL — should the kdc answer digest requests, or should it be mysterious and unbothered? default is FALSE.
  • digests_allowed = list of digests — tells you which digests the kdc will actually reply to. default is ntlm-v2.
  • kx509_ca = file — points at the PEM credentials for the kx509 certification authority. 🔐
  • require_initial_kca_tickets = boolean — says whether tickets for the kca_service service principal have to be INITIAL. you can set this per-realm or globally, your choice bestie. defaults to true for the global setting.
  • kx509_include_pkinit_san = boolean — if it's true, then the kx509 client principal's name and realm get tucked into an id-pkinit-san certificate extension. settable per-realm or globally. defaults to true for the global setting.
  • kx509_template = file — the PEM file with a template for the certs that'll be issued. you can interpolate these variables in the subject name using ${variable} syntax:
    • principal-name — the full name of the kx509 client principal.
    • principal-name-without-realm — the full name of the kx509 client principal, but minus the realm name.
    • principal-name-realm — the name of the client principal's realm.
heads up bestie: the kx509, kx509_template, kx509_include_pkinit_san, and require_initial_kca_tickets parameters can also be set on a per-realm basis. the more u know ✨

🎀 [kadmin] 🎀

  • password_lifetime = time — if a principal already has its password set for expiration, this is how long it stays valid after a change. the clock starts ticking ⏳
  • default_keys = keytypes... — for every entry in default_keys, it tries to parse it as a sequence of etype:salttype:salt syntax, like this lil thing:
[(des|des3|etype):](pw-salt|afs3-salt)[:string]

if etype is omitted, that means everything. and if string is omitted, that means the default salt string (for that principal and encryption type). also there are extra special values of keytypes:

  • v5 — the Kerberos 5 salt pw-salt
  • default_key_rules = {
    • globing-rule = keytypes... — a globbing rule that matches a principal, and when it's true, it uses the keytypes as specified, same format as [kadmin]default_keys .
  • }
  • prune-key-history = BOOL — when it's adding keys to the key history, it drops keys that are too old to match unexpired tickets (based on the principal's maximum ticket lifetime). ⚠️ real talk: if the KDC keystore gets compromised later, traffic that was protected with the discarded older keys may stay protected. also this keeps the HDB records for principals with key history from growing without bound, which is honestly a vibe. the default (backwards compatible) value is "false".
  • use_v4_salt = BOOL — when it's true, it's literally the same as
default_keys = des3:pw-salt v4

and it's only left in for backwards compatibility. it's giving legacy, not giving main character 🫠

💖 [password_quality] 💖

go check the Password quality assurance in the info documentation for more deets, bestie. but here's the lineup:

  • check_library = library-name — library name that contains the password check_function
  • check_function = function-name — function name for checking passwords in check_library
  • policy_libraries = library1 ... libraryN — list of libraries that can do password policy checks
  • policies = policy1 ... policyN — list of policy names to apply to the password. builtin policies include, among others, minimum-length, character-class, external-check.

✨ ENVIRONMENT ✨

  • KRB5_CONFIG — points to the configuration file to read. that's it. that's the whole env var. iconic behavior honestly 👑

🌸 FILES 🌸

  • /etc/krb5.conf — configuration file for Kerberos 5.

💖 EXAMPLES 💖

okay here's a whole mood board of a config, look at her go:

[libdefaults]
        default_realm = FOO.SE
        name_canon_rules = as-is:realm=FOO.SE
        name_canon_rules = qualify:domain=foo.se:realm=FOO.SE
        name_canon_rules = qualify:domain=bar.se:realm=FOO.SE
        name_canon_rules = nss
[domain_realm]
        .foo.se = FOO.SE
        .bar.se = FOO.SE
[realms]
        FOO.SE = {
                kdc = kerberos.foo.se
                default_domain = foo.se
        }
[logging]
        kdc = FILE:/var/heimdal/kdc.log
        kdc = SYSLOG:INFO
        default = SYSLOG:INFO:USER
[kadmin]
        default_key_rules = {
                */ppp@* = arcfour-hmac-md5:pw-salt
        }

🩺 DIAGNOSTICS 🩺

so here's the tea 😭 — because krb5.conf gets read and parsed by the krb5 library, programs don't really get a lot of opportunities to report parsing errors in any useful format. to help with that whole situation, there's a program called verify_krb5_conf that reads krb5.conf and tries to emit useful diagnostics from parsing errors.

but ngl, this program has no way of knowing what options are actually used, so it cannot warn you about unknown or misspelled ones. it's doing its best, bestie. be nice to her 💀

👀 SEE ALSO 👀

and that's the whole page, bestie, we ate and left no crumbs 💅✨ this one's from HEIMDAL, dated May 4, 2005, and it lives at KRB5.CONF(5) — go configure something gorgeous 🦋🎀💖

📦 where this page comes from