On 10/27/19 7:33 AM, Mimi Zohar wrote:
Other examples of trusted keyrings are: .ima, .evm, .platform,
.blacklist, .builtin_regdb_keys. Instead of defining a keyring
specific method of getting the keyring number, define a generic
method. For example, the userspace command "keyctl describe
%keyring:.builtin_trusted_keys" searches /proc/keys, but the kernel
shouldn't need to access /proc/keys.
"description" field in "struct key" is set to ".builtin_trusted_keys"
for Built-In Trusted Keys keyring.
Similarly, for other keyrings such as .ima, .evm, .blacklist,
.builtin_regdb_keys, etc.
> # measure keys on the BUILTIN and IMA keyrings into a different PCR
> measure func=KEYRING_CHECK keyring=".builtin_trusted_keys|.ima" pcr=11
With IMA policy set like above, the keyring to keyring number mapping
can be set at IMA policy load.
My earlier point about mapping the "keyring" to "keyring number" at
runtime (when the IMA hook is called) still needs to be done to know if
the given keyring is in the policy or not.
void ima_post_key_create_or_update(struct key *keyring, struct key *key,
unsigned long flags, bool create)
thanks,
-lakshmi
