> On Sep 9, 2020, at 11:40 AM, Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote:
>
> On 9/9/20 10:27 AM, Eric Snowberg wrote:
>> diff --git a/include/crypto/pkcs7.h b/include/crypto/pkcs7.h
>> index 38ec7f5f9041..d8f2e0fdfbf4 100644
>> --- a/include/crypto/pkcs7.h
>> +++ b/include/crypto/pkcs7.h
>> @@ -26,11 +26,19 @@ extern int pkcs7_get_content_data(const struct pkcs7_message *pkcs7,
>> const void **_data, size_t *_datalen,
>> size_t *_headerlen);
>>
>> +#ifdef CONFIG_PKCS7_MESSAGE_PARSER
>> /*
>> * pkcs7_trust.c
>> */
>> extern int pkcs7_validate_trust(struct pkcs7_message *pkcs7,
>> struct key *trust_keyring);
>> +#else
>> +static inline int pkcs7_validate_trust(struct pkcs7_message *pkcs7,
>> + struct key *trust_keyring)
>> +{
>> + return -ENOKEY;
>> +}
>> +#endif
>
> Just to be clear, you want to do the #else block when
> CONFIG_PKCS7_MESSAGE_PARSER=m. Is that correct?
>
> If so, it might be clearer to use
>
> #if IS_BUILTIN(CONFIG_PKCS7_MESSAGE_PARSER)
>
I just added this part to fix a build error when none of the
asymmetrical keys are defined within a config. I failed to notice
CONFIG_PKCS7_MESSAGE_PARSER could be configured to build as a module
too. The code I added that uses pkcs7_validate_trust is always
builtin. Taking this into account, please disregard this patch.
I will need to solve this a different way. Thanks for pointing this
out.
