Re: [PATCH v9 5/8] IMA: limit critical data measurement based on a label

From: Tushar Sugandhi <tusharsu@linux.microsoft.com>
To: Mimi Zohar <zohar@linux.ibm.com>,
	stephen.smalley.work@gmail.com, casey@schaufler-ca.com,
	agk@redhat.com, snitzer@redhat.com, gmazyland@gmail.com,
	paul@paul-moore.com
Cc: tyhicks@linux.microsoft.com, sashal@kernel.org,
	jmorris@namei.org, nramas@linux.microsoft.com,
	linux-integrity@vger.kernel.org, selinux@vger.kernel.org,
	linux-security-module@vger.kernel.org,
	linux-kernel@vger.kernel.org, dm-devel@redhat.com
Subject: Re: [PATCH v9 5/8] IMA: limit critical data measurement based on a label
Date: Tue, 5 Jan 2021 12:28:19 -0800	[thread overview]
Message-ID: <2c1d83b6-e344-28ea-e387-01a0febbe391@linux.microsoft.com> (raw)
In-Reply-To: <56db41c08d625b8143454a2e0aaaef3ea2927442.camel@linux.ibm.com>

On 2020-12-24 6:29 a.m., Mimi Zohar wrote:
> Hi Tushar,
> 
> On Sat, 2020-12-12 at 10:02 -0800, Tushar Sugandhi wrote:
>> System administrators should be able to limit which kernel subsystems
>> they want to measure the critical data for. To enable that, an IMA policy
>> condition to choose specific kernel subsystems is needed. This policy
>> condition would constrain the measurement of the critical data based on
>> a label for the given subsystems.
> 
> Restricting which kernel integrity critical data is measured is not
> only of interest to system administrators.   Why single them out?
> 
system administrators are usually responsible for system 
policies/configurations.They own modifications in the config files like
ima-policy. That's why we wanted to address them to begin with. But you
are correct. This is not only of interest to sysadmins. I will make the 
description more generic.

> Limiting which critical data is measured is based on a label, making it
> flexible.  In your use case scenario, you're grouping the label based
> on kernel subsystem, but is that really necessary?  In the broader
> picture, there could be cross subsystem critical data being measured
> based on a single label.
> 
> Please think about the broader picture and re-write the patch
> descirption more generically.
> 
Makes sense. Will make the patch description more generic.
>>
>> Add a new IMA policy condition - "data_source:=" to the IMA func
> 
> What is with "add"?  You're "adding support for" or "defining" a new
> policy condition.  Remove the single hyphen, as explained in 3/8.
> 
> Please replace "data_source" with something more generic (e.g. label).
> 
Sounds good. Would you prefer "label" or something else like "data_label"?

In the policy file the "label" looks logical and more generic than 
"data_label".
    measure func=CRITICAL_DATA label=selinux

For the time being, I will stick with "label", please let me know if you
prefer something else.

Thanks,
Tushar

> thanks,
> 
> Mimi
> 
>> CRITICAL_DATA to allow measurement of various kernel subsystems. This
>> policy condition would enable the system administrators to restrict the
>> measurement to the labels listed in "data_source:=".
>>
>> Limit the measurement to the labels that are specified in the IMA
>> policy - CRITICAL_DATA+"data_source:=". If "data_sources:=" is not
>> provided with the func CRITICAL_DATA, the data from all the
>> supported kernel subsystems is measured.
>>
>> Signed-off-by: Tushar Sugandhi <tusharsu@linux.microsoft.com>