Display the package version number

keyctl --version

This command prints the package version number and build date and exits:

$ keyctl --version
keyctl from keyutils-1.5.3 (Built 2011-08-24)

Query subsystem capabilities

keyctl supports
keyctl supports --raw
keyctl supports <cap>

This command can list the available capabilities:

$ keyctl supports
have_capabilities=0
have_persistent_keyrings=1
have_dh_compute=1
have_public_key=1

produce a raw hex dump of the capabilities list:

$ keyctl supports --raw
ff0f

or query a specific capability:

$ keyctl supports pkey
echo $?
0

which exits 0 if the capability is supported, 1 if it isn't and 3 if the name is not recognised. The capabilities supported are:

capabilities

The kernel supports capability querying. If not, the other capabilities will be queried as best libkeyutils can manage.

persistent_keyrings

The kernel supports persistent keyrings.

dh_compute

The kernel supports Diffie-Hellman computation operations.

public_key

The kernel supports public key operations.

big_key_type

The kernel supports the big_key key type.

key_invalidate

The kernel supports the invalidate key operaiton.

restrict_keyring

The kernel supports the restrict_keyring operation.

move_key

The kernel supports the move key operation.

ns_keyring_name

Keyring names are segregated according to the user-namespace in which the keyrings are created.

ns_key_tag

Keys can get tagged with namespace tags, allowing keys with the same type and description, but different namespaces to coexist in the same keyring. Tagging is done automatically according to the key type.

Show actual key or keyring ID

keyctl id [<key>]

This command looks up the real ID of a key or keyring from the identifier given, which is typically a symbolic ID such as "@s" indicating the session keyring, but can also be a numeric ID or "%type:desc" notation. If a special keyring is specified that isn't created yet, an error will be given rather than causing that keyring to be created.

Show process keyrings

keyctl show [-x] [<keyring>]

By default this command recursively shows what keyrings a process is subscribed to and what keys and keyrings they contain. If a keyring is specified then that keyring will be dumped instead. If -x is specified then the keyring IDs will be dumped in hex instead of decimal.

Add a key to a keyring

keyctl add [-x] <type> <desc> <data> <keyring>
keyctl padd [-x] <type> <desc> <keyring>

This command creates a key of the specified type and description; instantiates it with the given data and attaches it to the specified keyring. It then prints the new key's ID on stdout:

$ keyctl add user mykey stuff @u
26

The padd variant of the command reads the data from stdin rather than taking it from the command line:

If -x is given, then the data is hex-decoded with whitespace being discarded.

Request a key

keyctl request <type> <desc> [<dest_keyring>]
keyctl request2 <type> <desc> <info> [<dest_keyring>]
keyctl prequest2 <type> <desc> [<dest_keyring>]

These three commands request the lookup of a key of the given type and description. The process's keyrings will be searched, and if a match is found the matching key's ID will be printed to stdout; and if a destination keyring is given, the key will be added to that keyring also.

If there is no key, the first command will simply return the error ENOKEY and fail. The second and third commands will create a partial key with the type and description, and call out to /sbin/request-key with that key and the extra information supplied. This will then attempt to instantiate the key in some manner, such that a valid key is obtained.

The third command is like the second, except that the callout information is read from stdin rather than being passed on the command line.

If a valid key is obtained, the ID will be printed and the key attached as if the original search had succeeded.

If there wasn't a valid key obtained, a temporary negative key will be attached to the destination keyring if given and the error "Requested key not available" will be given.

$ keyctl request2 user debug:hello wibble
23
$ echo -n wibble | keyctl prequest2 user debug:hello
23
$ keyctl request user debug:hello
23

Update a key

keyctl update [-x] <key> <data>
keyctl pupdate [-x] <key>

This command replaces the data attached to a key with a new set of data. If the type of the key doesn't support update then error "Operation not supported" will be returned.

$ keyctl update 23 zebra

The pupdate variant of the command reads the data from stdin rather than taking it from the command line:

$ echo -n zebra | keyctl pupdate 23
$ echo 616263313233 | keyctl pupdate -x 23

If -x is given, then the data is hex-decoded with whitespace being discarded.

Create a keyring

keyctl newring <name> <keyring>

This command creates a new keyring of the specified name and attaches it to the specified keyring. The ID of the new keyring will be printed to stdout if successful.

$ keyctl newring squelch @us
27

Revoke a key

keyctl revoke <key>

This command marks a key as being revoked. Any further operations on that key (apart from unlinking it) will return error "Key has been revoked".

$ keyctl revoke 26
$ keyctl describe 26
keyctl_describe: Key has been revoked

Clear a keyring

keyctl clear <keyring>

This command unlinks all the keys attached to the specified keyring. Error "Not a directory" will be returned if the key specified is not a keyring.

$ keyctl clear 27

Link a key to a keyring

keyctl link <key> <keyring>

This command makes a link from the key to the keyring if there's enough capacity to do so. Error "Not a directory" will be returned if the destination is not a keyring. Error "Permission denied" will be returned if the key doesn't have link permission or the keyring doesn't have write permission. Error "File table overflow" will be returned if the keyring is full. Error "Resource deadlock avoided" will be returned if an attempt was made to introduce a recursive link.

$ keyctl link 23 27
$ keyctl link 27 27
keyctl_link: Resource deadlock avoided

Unlink a key from a keyring or the session keyring tree

keyctl unlink <key> [<keyring>]

If the keyring is specified, this command removes a link to the key from the keyring. Error "Not a directory" will be returned if the destination is not a keyring. Error "Permission denied" will be returned if the keyring doesn't have write permission. Error "No such file or directory" will be returned if the key is not linked to by the keyring.

If the keyring is not specified, this command performs a depth-first search of the session keyring tree and removes all the links to the nominated key that it finds (and that it is permitted to remove). It prints the number of successful unlinks before exiting.

$ keyctl unlink 23 27

Move a key between keyrings.

keyctl move [-f] <key> <from_keyring> <to_keyring>

This command moves a key from one keyring to another, atomically combining "keyctl unlink <key> <from_keyring>" and "keyctl link <key> <to_keyring>".

If the "-f" flag is present, any matching key will be displaced from "to_keyring"; if not present, the command will fail with the error message "File exists" if the key would otherwise displace another key from "to_keyring".

$ keyctl move 23 27 29
$ keyctl move -f 71 @u @s

Search a keyring

keyctl search <keyring> <type> <desc> [<dest_keyring>]

This command non-recursively searches a keyring for a key of a particular type and description. If found, the ID of the key will be printed on stdout and the key will be attached to the destination keyring if present. Error "Requested key not available" will be returned if the key is not found.

$ keyctl search @us user debug:hello
23
$ keyctl search @us user debug:bye
keyctl_search: Requested key not available

Restrict a keyring

keyctl restrict_keyring <keyring> [<type> [<restriction>]]

This command limits the linkage of keys to the given keyring using a provided restriction scheme. The scheme is associated with a given key type, with further details provided in the restriction option string. Options typically contain a restriction name possibly followed by key ids or other data relevant to the restriction. If no restriction scheme is provided, the keyring will reject all links.

$ keyctl restrict_keyring $1 asymmetric builtin_trusted

Read a key

keyctl read <key>
keyctl pipe <key>
keyctl print <key>

These commands read the payload of a key. "read" prints it on stdout as a hex dump, "pipe" dumps the raw data to stdout and "print" dumps it to stdout directly if it's entirely printable or as a hexdump preceded by ":hex:" if not.

If the key type does not support reading of the payload, then error "Operation not supported" will be returned.

$ keyctl read 26
1 bytes of data in key:
62
$ keyctl print 26
b
$ keyctl pipe 26
$

List a keyring

keyctl list <keyring>
keyctl rlist <keyring>

These commands list the contents of a key as a keyring. "list" pretty prints the contents and "rlist" just produces a space-separated list of key IDs.

No attempt is made to check that the specified keyring is a keyring.

$ keyctl list @us
2 keys in keyring:


       22: vrwsl----------  4043    -1 keyring: _uid.4043


       23: vrwsl----------  4043  4043 user: debug:hello
$ keyctl rlist @us
22 23

Describe a key

keyctl describe <keyring>
keyctl rdescribe <keyring> [sep]

These commands fetch a description of a keyring. "describe" pretty prints the description in the same fashion as the "list" command; "rdescribe" prints the raw data returned from the kernel.

$ keyctl describe @us


       -5: vrwsl----------  4043    -1 keyring: _uid_ses.4043
$ keyctl rdescribe @us
keyring;4043;-1;3f1f0000;_uid_ses.4043

The raw string is "<type>;<uid>;<gid>;<perms>;<description>", where uid and gid are the decimal user and group IDs, perms is the permissions mask in hex, type and description are the type name and description strings (neither of which will contain semicolons).

Change the access controls on a key

keyctl chown <key> <uid>
keyctl chgrp <key> <gid>

These two commands change the UID and GID associated with evaluating a key's permissions mask. The UID also governs which quota a key is taken out of.

The chown command is not currently supported; attempting it will earn the error "Operation not supported" at best.

For non-superuser users, the GID may only be set to the process's GID or a GID in the process's groups list. The superuser may set any GID it likes.

$ sudo keyctl chown 27 0
keyctl_chown: Operation not supported
$ sudo keyctl chgrp 27 0

Set the permissions mask on a key

keyctl setperm <key> <mask>

This command changes the permission control mask on a key. The mask may be specified as a hex number if it begins "0x", an octal number if it begins "0" or a decimal number otherwise.

The hex numbers are a combination of:

Possessor UID       GID       Other     Permission Granted
========  ========  ========  ========  ==================
01000000  00010000  00000100  00000001  View
02000000  00020000  00000200  00000002  Read
04000000  00040000  00000400  00000004  Write
08000000  00080000  00000800  00000008  Search
10000000  00100000  00001000  00000010  Link
20000000  00200000  00002000  00000020  Set Attribute
3f000000  003f0000  00003f00  0000003f  All

View permits the type, description and other parameters of a key to be viewed.

Read permits the payload (or keyring list) to be read if supported by the type.

Write permits the payload (or keyring list) to be modified or updated.

Search on a key permits it to be found when a keyring to which it is linked is searched.

Link permits a key to be linked to a keyring.

Set Attribute permits a key to have its owner, group membership, permissions mask and timeout changed.

$ keyctl setperm 27 0x1f1f1f00

Start a new session with fresh keyrings

keyctl session
keyctl session - [<prog> <arg1> <arg2> ...]
keyctl session <name> [<prog> <arg1> <arg2> ...]

These commands join or create a new keyring and then run a shell or other program with that keyring as the session key.

The variation with no arguments just creates an anonymous session keyring and attaches that as the session keyring; it then exec's $SHELL.

The variation with a dash in place of a name creates an anonymous session keyring and attaches that as the session keyring; it then exec's the supplied command, or $SHELL if one isn't supplied.

The variation with a name supplied creates or joins the named keyring and attaches that as the session keyring; it then exec's the supplied command, or $SHELL if one isn't supplied.

$ keyctl rdescribe @s
keyring;4043;-1;3f1f0000;_uid_ses.4043
$ keyctl session
Joined session keyring: 28
$ keyctl rdescribe @s
keyring;4043;4043;3f1f0000;_ses.24082
$ keyctl session -
Joined session keyring: 29
$ keyctl rdescribe @s
keyring;4043;4043;3f1f0000;_ses.24139
$ keyctl session - keyctl rdescribe @s
Joined session keyring: 30
keyring;4043;4043;3f1f0000;_ses.24185
$ keyctl session fish
Joined session keyring: 34
$ keyctl rdescribe @s
keyring;4043;4043;3f1f0000;fish
$ keyctl session fish keyctl rdesc @s
Joined session keyring: 35
keyring;4043;4043;3f1f0000;fish

Instantiate a key

keyctl instantiate [-x] <key> <data> <keyring>
keyctl pinstantiate [-x] <key> <keyring>
keyctl negate <key> <timeout> <keyring>
keyctl reject <key> <timeout> <error> <keyring>

These commands are used to attach data to a partially set up key (as created by the kernel and passed to /sbin/request-key). "instantiate" marks a key as being valid and attaches the data as the payload. "negate" and "reject" mark a key as invalid and sets a timeout on it so that it'll go away after a while. This prevents a lot of quickly sequential requests from slowing the system down overmuch when they all fail, as all subsequent requests will then fail with error "Requested key not found" (if negated) or the specified error (if rejected) until the negative key has expired.

Reject's error argument can either be a UNIX error number or one of 'rejected', 'expired' or 'revoked'.

The newly instantiated key will be attached to the specified keyring.

These commands may only be run from the program run by request-key - a special authorisation key is set up by the kernel and attached to the request-key's session keyring. This special key is revoked once the key to which it refers has been instantiated one way or another.

$ keyctl instantiate $1 "Debug $3" $4
$ keyctl negate $1 30 $4
$ keyctl reject $1 30 64 $4

The pinstantiate variant of the command reads the data from stdin rather than taking it from the command line:

$ echo -n "Debug $3" | keyctl pinstantiate $1 $4

If -x is given, then the data is hex-decoded with whitespace being discarded:

$ echo 01 02 03 04 | keyctl pinstantiate -x $1 $4

Set the expiry time on a key

keyctl timeout <key> <timeout>

This command is used to set the timeout on a key, or clear an existing timeout if the value specified is zero. The timeout is given as a number of seconds into the future.

$ keyctl timeout $1 45

Retrieve a key's security context

keyctl security <key>

This command is used to retrieve a key's LSM security context. The label is printed on stdout.

$ keyctl security @s
unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023

Give the parent process a new session keyring

keyctl new_session [<name>]

This command is used to give the invoking process (typically a shell) a new session keyring, discarding its old session keyring. If a name is given, the keyring is given that name, otherwise it will be given a name of "_ses" and will not be manually joinable.

$  keyctl session foo
Joined session keyring: 723488146
$  keyctl show
Session Keyring


       -3 --alswrv      0     0  keyring: foo
$  keyctl new_session
490511412
$  keyctl show
Session Keyring


       -3 --alswrv      0     0  keyring: _ses

Note that this affects the parent of the process that invokes the system call, and so may only affect processes with matching credentials. Furthermore, the change does not take effect till the parent process next transitions from kernel space to user space - typically when the wait() system call returns.

Remove dead keys from the session keyring tree

keyctl reap

This command performs a depth-first search of the caller's session keyring tree and attempts to unlink any key that it finds that is inaccessible due to expiry, revocation, rejection or negation. It does not attempt to remove live keys that are unavailable simply due to a lack of granted permission.

A key that is designated reapable will only be removed from a keyring if the caller has Write permission on that keyring, and only keyrings that grant Search permission to the caller will be searched.

The command prints the number of keys reaped before it exits. If the -v flag is passed then the reaped keys are listed as they're being reaped, together with the success or failure of the unlink.

Remove matching keys from the session keyring tree

keyctl purge <type>
keyctl purge [-i] [-p] <type> <desc>
keyctl purge -s <type> <desc>

These commands perform a depth-first search to find matching keys in the caller's session keyring tree and attempts to unlink them. The number of keys successfully unlinked is printed at the end.

The keyrings must grant Read and View permission to the caller to be searched, and the keys to be removed must also grant View permission. Keys can only be removed from keyrings that grant Write permission.

The first variant purges all keys of the specified type.

The second variant purges all keys of the specified type that also match the given description literally. The -i flag allows a case-independent match and the -p flag allows a prefix match.

The third variant purges all keys of the specified type and matching description using the key type's comparator in the kernel to match the description. This permits the key type to match a key with a variety of descriptions.

Get persistent keyring

keyctl get_persistent <keyring> [<uid>]

This command gets the persistent keyring for either the current UID or the specified UID and attaches it to the nominated keyring. The persistent keyring's ID will be printed on stdout.

The kernel will create the keyring if it doesn't exist and every time this command is called, will reset the expiration timeout on the keyring to the value in:

/proc/sys/kernel/keys/persistent_keyring_expiry

(by default three days). Should the timeout be reached, the persistent keyring will be removed and everything it pins can then be garbage collected.

If a UID other than the process's real or effective UIDs is specified, then an error will be given if the process does not have the CAP_SETUID capability.

Compute a Diffie-Hellman shared secret or public key

keyctl dh_compute <private> <prime> <base>

This command computes either a Diffie-Hellman shared secret or the public key corresponding to the provided private key using the payloads of three keys. The computation is:

base ^ private (mod prime)

The three inputs must be user keys with read permission. If the provided base key contains the shared generator value, the public key will be computed. If the provided base key contains the remote public key value, the shared secret will be computed.

The result is printed to stdout as a hex dump.

$ keyctl dh_compute $1 $2 $3
8 bytes of data in result:
00010203 04050607

Compute a Diffie-Hellman shared secret and derive key material

keyctl dh_compute_kdf <private> <prime> <base> <output_length> <hash_type>

This command computes a Diffie-Hellman shared secret and derives key material from the shared secret using a key derivation function (KDF). The shared secret is derived as outlined above and is input to the KDF using the specified hash type. The hash type must point to a hash name known to the kernel crypto API.

The operation derives key material of the length specified by the caller.

The operation is compliant to the specification of SP800-56A.

The result is printed to stdout as hex dump.

Compute a Diffie-Hellman shared secret and apply KDF with other input

keyctl dh_compute_kdf_oi [-x] <private> <prime> <base> <output_length> <hash_type>

This command is identical to the command dh_compute_kdf to generate a Diffie-Hellman shared secret followed by a key derivation operation. This command allows the caller to provide the other input data (OI data) compliant to SP800-56A via stdin.

If -x is given, then the data passed to stdin is hex-decoded with whitespace being discarded.

Perform public-key operations with an asymmetric key

keyctl pkey_query <key> <pass> [k=v]*
keyctl pkey_encrypt <key> <pass> <datafile> [k=v]* > <encfile>
keyctl pkey_decrypt <key> <pass> <encfile> [k=v]* > <datafile>
keyctl pkey_sign <key> <pass> <datafile> [k=v]* > <sigfile>
keyctl pkey_verify <key> <pass> <datafile> <sigfile> [k=v]*

These commands query an asymmetric key, encrypt data with it, decrypt the encrypted data, generate a signature over some data and verify that signature. For encrypt, decrypt and sign, the resulting data is written to stdout; verify reads the data and the signature files and compares them.

[!] NOTE that the data is of very limited capacity, with no more bits than the size of the key. For signatures, the caller is expected to digest the actual data and pass in the result of the digest as the datafile. The name of the digest should be specified on the end of the command line as "hash=<name>".

The key ID indicates the key to use; pass is a placeholder for future password provision and should be "0" for the moment; datafile is the unencrypted data to be encrypted, signed or to have its signature checked; encfile is a file containing encrypted data; and sigfile is a file containing a signature.

A list of parameters in "key[=val]" form can be included on the end of the command line. These specify things like the digest algorithm used ("hash=<name>") or the encoding form ("enc=<type>").

k=`keyctl padd asymmetric "" @s <key.pkcs8.der`
keyctl pkey_query $k 0 enc=pkcs1 hash=sha256
keyctl pkey_encrypt $k 0 foo.hash enc=pkcs1 >foo.enc
keyctl pkey_decrypt $k 0 foo.enc enc=pkcs1 >foo.hash
keyctl pkey_sign $k 0 foo.hash enc=pkcs1 hash=sha256 >foo.sig
keyctl pkey_verify $k 0 foo.hash foo.sig enc=pkcs1 hash=sha256

See asymmetric-key(7) for more information.

Change notifications

keyctl watch [-f<filters>] <key>
keyctl watch_session [-f <filters>] [-n <name>] \
<notifylog> <gclog> <fd> <prog> [<arg1> <arg2> ...] keyctl watch_add <fd> <key>
keyctl watch_rm <fd> <key>

The watch command watches a single key, printing notifications to stdout until the key is destroyed. Filters can be employed to cut down the events that will be delivered. The filter string is a series of letters, each one of which enables a particular event subtype:

i - The key has been instantiated
p - The key has been updated
l - A link has been added to a keyring
n - A link has been removed from a keyring
c - A keyring has been cleared
r - A key has been revoked
v - A key has been invalidated
s - A key has had its attributes changed

The output of the command looks like:

<keyid> <event> [<aux>]

Where keyid is the primary subject of the notification, op is the event and aux is the secondary key if there is one (such as link where the primary key is the keyring secondary key is the key being linked in to it). For example:

255913279 link 340681059
255913279 clr

An additional notication is generated when a key being watched is garbage collected, e.g.:

255913279 gc

The watch_session command creates a new session keyring, with name name if given, watches it for notifications and runs program prog with it. The program is given the specified arguments.

A second process is forked off to monitor the notifications. The output from that is directed to the files notifylog for most notifications and gclog for key removal notifications (which are asynchronous and may be deferred).

The watch_queue(7) device is exported to the program attached to fd number fd. This can be passed by the other two commands.

The watch_add command adds a watch on key to the watch_queue attached to fd as exported by watch_session and the watch_rm caommand removes it. A watch_queue can handle multiple keys and even non-keys sources as well.