Re: linux-next: Tree for Apr 16 (IMA appraise causing build error)

[Randy Dunlap]

Hi,

On 4/16/21 1:25 PM, Nayna wrote:
> 
> On 4/16/21 2:53 PM, Randy Dunlap wrote:
>> On 4/16/21 4:36 AM, Stephen Rothwell wrote:
>>> Hi all,
>>>
>>> Changes since 20210415:
>>>
>> I noticed this build error message (on an i386 build):
>>
>> ../certs/Makefile:52: *** Could not determine digest type to use from kernel config.  Stop.
>>
>> and when I was checking on why it happened, I noticed that
>> # CONFIG_MODULES is not set
>>
>> and hence
>> ifndef CONFIG_MODULE_SIG_HASH
>> $(error Could not determine digest type to use from kernel config)
>> endif
>>
>> CONFIG_MODULE_SIG_HASH is not set/enabled/defined.
>>
>> However, the .config file does have
>> CONFIG_IMA_APPRAISE=y
>> # CONFIG_IMA_ARCH_POLICY is not set
>> # CONFIG_IMA_APPRAISE_BUILD_POLICY is not set
>> CONFIG_IMA_APPRAISE_BOOTPARAM=y
>> CONFIG_IMA_APPRAISE_MODSIG=y
>>
>> as well as
>> CONFIG_MODULE_SIG_FORMAT=y
>>
>> due to a "select" by IMA_APPRAISE_MODSIG.
>> (although I see that MODULE_SIG_FORMAT does not depend on MODULES)
>>
>>
>> Is there anything that you can do (or recommend) to prevent
>> the build error?
>>
>>
>>
>> BTW, it looks like this:
>> config IMA_APPRAISE_REQUIRE_MODULE_SIGS
>>     bool "Appraise kernel modules signatures"
>>     depends on IMA_APPRAISE_BUILD_POLICY
>>
>> could also depend on MODULES.
>>
>>
>>
>> Full i386 randconfig file is attached.
> 
> 
> With the new patchset "ima: kernel build support for loading the kernel module signing key", there shouldn't be a difference when generating the config file between MODULE_SIG and IMA_APPRAISE_MODSIG. Both prompt for the hash algorithm.

That patchset appears to be included in today's linux-next 2021-04-16.

> Can you please explain how you generate randconfig? Do you use make xconfig?

with the 'make randconfig' command.


-- 
~Randy