The key identifiers passed to or returned from keyctl are, in general, positive integers. There are, however, some special values with special meanings that can be passed as arguments:
(*) No key: 0
(*) Thread keyring: @t or -1
Each thread may have its own keyring. This is searched first, before all others. The thread keyring is replaced by (v)fork, exec and clone.
(*) Process keyring: @p or -2
Each process (thread group) may have its own keyring. This is shared between all members of a group and will be searched after the thread keyring. The process keyring is replaced by (v)fork and exec.
(*) Session keyring: @s or -3
Each process subscribes to a session keyring that is inherited across (v)fork, exec and clone. This is searched after the process keyring. Session keyrings can be named and an extant keyring can be joined in place of a process's current session keyring.
(*) User specific keyring: @u or -4
This keyring is shared between all the processes owned by a particular user. It isn't searched directly, but is normally linked to from the session keyring.
(*) User default session keyring: @us or -5
This is the default session keyring for a particular user. Login processes that change to a particular user will bind to this session until another session is set.
(*) Group specific keyring: @g or -6
This is a place holder for a group specific keyring, but is not actually implemented yet in the kernel.
(*) Assumed request_key authorisation key: @a or -7
This selects the authorisation key provided to the request_key() helper to permit it to access the callers keyrings and instantiate the target key.
(*) Keyring by name: %:<name>
A named keyring. This will be searched for in the process's keyrings and in /proc/keys.
(*) Key by name: %<type>:<name>
A named key of the given type. This will be searched for in the process's keyrings and in /proc/keys.
(*) Display the package version number
This command prints the package version number and build date and exits:
(*) 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 <type> <desc> <data> <keyring>
keyctl padd <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:
The padd variant of the command reads the data from stdin rather than taking it from the command line:
(*) 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.
(*) Update a key
keyctl update <key> <data>
keyctl pupdate <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.
The pupdate variant of the command reads the data from stdin rather than taking it from the command line:
(*) 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.
(*) 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".
(*) 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.
(*) 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.
(*) 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.
(*) 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.
(*) 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.
(*) 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.
(*) 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.
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.
(*) 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:
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.
(*) Start a new session with fresh keyrings
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.
Joined session keyring: 28
testbox>keyctl rdescribe @s
testbox>keyctl session -
Joined session keyring: 29
testbox>keyctl rdescribe @s
testbox>keyctl session - keyctl rdescribe @s
Joined session keyring: 30
testbox>keyctl session fish
Joined session keyring: 34
testbox>keyctl rdescribe @s
testbox>keyctl session fish keyctl rdesc @s
Joined session keyring: 35
(*) Instantiate a key
keyctl instantiate <key> <data> <keyring>
keyctl pinstantiate <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.
The pinstantiate variant of the command reads the data from stdin rather than taking it from the command line:
(*) 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.
(*) 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.
(*) Give the parent process a new session keyring
This command is used to give the invoking process (typically a shell) a new session keyring, discarding its old session keyring.
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
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:
(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.
There are a number of common errors returned by this program:
"Not a directory" - a key wasn't a keyring.
"Requested key not found" - the looked for key isn't available.
"Key has been revoked" - a revoked key was accessed.
"Key has expired" - an expired key was accessed.
"Permission denied" - permission was denied by a UID/GID/mask combination.