[PATCH v10 26/27] ima: Limit number of policy rules in non-init_ima_ns

[Stefan Berger]

On 2/23/22 12:04, Mimi Zohar wrote:
> On Wed, 2022-02-23 at 11:37 -0500, Stefan Berger wrote:
>> On 2/23/22 10:38, Mimi Zohar wrote:
>>> On Tue, 2022-02-01 at 15:37 -0500, Stefan Berger wrote:
>>>> Limit the number of policy rules one can set in non-init_ima_ns to a
>>>> hardcoded 1024 rules. If the user attempts to exceed this limit by
>>>> setting too many additional rules, emit an audit message with the cause
>>>> 'too-many-rules' and simply ignore the newly added rules.
>>> This paragraph describes 'what' you're doing, not 'why'.  Please prefix
>>> this paragraph with a short, probably one sentence, reason for the
>>> change.
>>>> Switch the accounting for the memory allocated for IMA policy rules to
>>>> GFP_KERNEL_ACCOUNT so that cgroups kernel memory accounting takes effect.
>>> Does this change affect the existing IMA policy rules for init_ima_ns?
>> There's typically no cgroup for the int_ima_ns, so not effect on
>> init_ima_ns.
>>
>> Here's the updated text:
>>
>> Limit the number of policy rules one can set in non-init_ima_ns to a
>> hardcoded 1024 rules to restrict the amount of memory used for IMA's
>> policy.
> The question is "why" there should be a difference between the
> init_ima_ns and non-init_ima_ns cases.  Perhaps something like, "Only


Chistian had suggested it to not have the number of policy rules unbounded.


> host root with CAP_SYS_ADMIN may write init_ima_ns policy rules, but in
> the non-init_ima_ns case root in the namespace with CAP_MAC_ADMIN
> privileges may write IMA policy rules.  Limit the total number of IMA
> policy rules per namespace."

What does it have to do with CAP_SYS_ADMIN and CAP_MAC_ADMIN and why 
mention these here? It seem primarily a limit of number of rules to 
avoid huge kernel memory consumption in the case that a cgroup limit for 
memory was not set up.


>
>>   Ignore the added rules if the user attempts to exceed this
>> limit by setting too many additional rules.
>>
>> Switch the accounting for the memory allocated for IMA policy rules to
>> GFP_KERNEL_ACCOUNT so that cgroups kernel memory accounting takes effect.
>> This switch has no effect on the init_ima_ns.
> thanks,
>
> Mimi
>