linux-kernel - Re: [PATCH v4 08/10] IMA: Defined functions to queue and dequeue keys for measurement

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <8b2fd578-7429-f5b8-4286-1face91e1ae6@linux.microsoft.com>
Date:   Wed, 6 Nov 2019 15:52:31 -0800
From:   Lakshmi Ramasubramanian <nramas@...ux.microsoft.com>
To:     Mimi Zohar <zohar@...ux.ibm.com>, dhowells@...hat.com,
        matthewgarrett@...gle.com, sashal@...nel.org,
        jamorris@...ux.microsoft.com, linux-integrity@...r.kernel.org,
        linux-security-module@...r.kernel.org, keyrings@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v4 08/10] IMA: Defined functions to queue and dequeue keys
 for measurement

On 11/6/2019 2:44 PM, Mimi Zohar wrote:

>> +int ima_queue_or_process_key_for_measurement(struct key *keyring,
>> +					     struct key *key)
>> +{
>> +	int rc = 0;
>> +	struct ima_measure_key_entry *entry = NULL;
>> +	const struct public_key *pk;
>> +
>> +	if (key->type != &key_type_asymmetric)
>> +		return 0;
>> +
>> +	mutex_lock(&ima_measure_keys_mutex);

> 
> Unless the key is being queued, there's no reason to take the lock.

Reason the lock is taken even in the case the key is not queued is to 
avoid the following race condition:

  => ima_init() sets ima_initialized flag and calls the dequeue function

  => If IMA hook checks ima_initialized flag outside the lock and sees 
the flag is not set, it will call the queue function.

  => If the above two steps race, the key could get added to the queue 
after ima_init() has processed the queued keys.

That's the reason I named the function called by the IMA hook to 
ima_queue_or_process_key_for_measurement().

But I can make the following change:

  => IMA hook checks the flag.
  => If it is set, process key immediately
  => If the flag is not set, call ima_queue_or_process_key_for_measurement()

ima_queue_or_process_key_for_measurement() will do the following:

  => With the lock held check ima_initialized flag
  => If true release the lock and call process_buffer_measurement()
  => If false, queue the key and then release the lock

Would that be acceptable?

> Measuring the key should be done in ima_post_key_create_or_update()
> unless, it is being deferred.  Please update the function name to
> reflect this.

Just wanted to confirm:
Rename ima_post_key_create_or_update() to a more appropriate name?

Another reason for doing all key related operations in 
ima_queue_or_process_key_for_measurement() is to isolate key related 
code in a separate C file and build it if and only if the CONFIG 
dependencies are met.

With respect to loading custom policy, I will take a look at how to 
handle that case. Thanks for pointing that out.

> Mimi

thanks,
  -lakshmi