Re: Creating a PKCS7 detached signature with providers

 

Hello,

I found the source of the problem and the solution, although it's not very 
elegant and there may be better ways to do it.

My goal is to create a detached PKCS7 signature, so I'm keeping the 
PKCS7_sign function and I have the following code (simplified, without 
error handling):

/* Create my own provider */
EVP_MD *md=EVP_MD_fetch(NULL, "SHA256", NULL) ;
/* Modern providers do not have this field, but it is essential for the 
signature to work. We add it in a dirty way for compatibility. */
*md->type=NID_sha256 ;*
/* Like PKCS7_sign_ex but lighter */ 
PKCS7 *pP7=PKCS7_new_ex(NULL, NULL);
PKCS7_set_type(pP7, NID_pkcs7_signed) ;
PKCS7_content_new(pP7, NID_pkcs7_data) ;
/* Adding information */
PKCS7_sign_add_signer(pP7, PX509Cert, pPrivKey, md, PKCS7_NOATTR);
PKCS7_set_detached(pP7, 1) ;
/* Signature */
PKCS7_final(pP7, pDataBIOIn, NULL, PKCS7_BINARY);

 

The problem is caused by this call stack:
*EVP_MD_get_type*
PKCS7_SIGNER_INFO_set
PKCS7_add_signature
PKCS7_sign_add_signer

Since an OpenSSL 3.0+ provider does not have a “type” field when fetched, 
the ‘EVP_MD_get_type’ function always returns 0 and corrupts all subsequent 
processing because "X509_ALGOR_set0(..., OBJ_nid2obj(EVP_MD_get_type(...)), 
...)“ does not set the ”digest_alg" field of my PKCS7 structure.


Le vendredi 3 octobre 2025 à 15:07:09 UTC+2, Mad a écrit :

> Thank you for your reply and sorry for the length of mine, but I think I 
> need to be exhaustive.
>
> For more detail, I've tried quite a few things to make myself compatible. 
> The executed code looks like this (I don't include error handling so as not 
> to make the code too heavy but I handle each function return correctly):
>
> EVP_MD *md=EVP_MD_fetch(NULL, "SHA256", NULL);
> CMS_ContentInfo *pCMS=CMS_sign(NULL, NULL, NULL, NULL, CMS_PARTIAL | 
> CMS_DETACHED);
> CMS_add1_signer(pCMS, PX509Cert, pPrivKey, md, CMS_NOATTR);
> CMS_final(pCMS, pDataBIOIn, NULL, SMIME_BINARY);
>
> If I don't prepare anything before, when I call ‘CMS_final’ I get the 
> following error "lib(46): ossl_cms_DigestAlgorithm_init_bio(149)‘ which 
> corresponds to ’ERR_raise(ERR_LIB_CMS, CMS_R_UNKOWN_DIGEST_ALGORITHM);".
>
> Of the attempts I have tried, the one that seems the least bad was to add 
> ‘EVP_add_digest(EVP_MD_fetch(NULL, “SHA256”, NULL));’ after initialising my 
> providers. In this case ‘CMS_final’ does not cause an error, but when I 
> retrieve my signature via ‘cbP7Data=id2_CMS_ContentInfo(pCMS, NULL);’ I get 
> ‘cbP7Data==-1’ and the error "lib(13) : asn1_template_ex :reason(222)' 
> which corresponds to 'ERR_raise(ERR_LIB_ASN1, 
> ASN1_R_ILLEGAL_ZERO_CONTENT);" but I can't find out which one it 
> corresponds to.
> I have the impression that something is missing when calling ‘CMS_final’ 
> but I can't say what.
>
> With regard to the digest providers that I define, I implement 
> PROV_NAMES_SHA1, PROV_NAMES_SHA256, PROV_NAMES_512 and PROV_NAMES_MD5.
>
> Finally, I'm not sure this is very useful but, as I'm in a restricted 
> environment, I had to limit as much as possible the modules compiled with 
> these options when generating my OpenSSL lib (it's not impossible that I 
> removed something too much but I don't know what):
>
> « no-autoalginit no-afalgeng no-aria no-asan no-asm no-async 
> no-autoload-config no-bf no-blake2 no-camellia no-capieng no-cast no-chacha 
> no-cmac no-cmp no-crypto-mdebug-backtrace no-ct no-deprecated no-des 
> no-dgram no-dh no-dsa no-ec no-dynamic-engine no-ec no-ec2m no-ecdh 
> no-ecdsa no-err no-external-tests no-filenames no-fuzz-libfuzzer 
> no-fuzz-afl no-gost no-heartbeats no-hw no-idea no-makedepend no-md2 no-md4 
> no-mdc2 no-msan no-multiblock no-nextprotoneg no-pinshared no-ocb no-ocsp 
> no-pic no-posix-io no-psk no-poly1305 no-rc2 no-rc4 no-rc5 no-rdrand 
> no-rfc3779 no-rmd160 no-scrypt no-siphash no-sm2 no-sm3 no-sm4 no-srp 
> no-srtp no-sse2 no-ssl no-ssl-trace no-static-engine no-stdio no-tests 
> no-threads no-ts no-ui no-ubsan no-ui-console no-unit-test no-winstore 
> no-whirlpool no-comp no-sctp no-seed no-shared no-ec_nistp_64_gcc_128 
> no-egd no-dsa no-dso no-dtls no-dtls1 no-dtls1_2 no-legacy no-sock no-tls 
> no-tls1 no-tls1_1 no-tls1_2 no-tls1_3 »
>
> Mad
>
>
> Le jeudi 2 octobre 2025 à 18:13:16 UTC+2, Matt Caswell a écrit :
>
>> On Thu, 2 Oct 2025 at 10:40, Mad <[email protected]> wrote:
>>
>>> Hello,
>>>
>>> This is the first time I've posted on a mailing list, so sorry if the 
>>> tone isn't correct.
>>>
>>> I'm in the process of finalising a project to update OpenSSL 1.1.1 to 
>>> 3.5.X with the purpose of doing ASN1 decoding and PKCS7 signing where I 
>>> integrate my own cryptography via providers (digest, cipher, mac, 
>>> signature, asym_cipher, keymgmt and kdf).
>>>
>>> When implementing the detached PKCS7 signature, I tried to use 
>>> ‘PKCS7_sign’ or ‘CMS_sign(NULL, NULL, NULL, NULL, CMS_PARTIAL|CMS_DETACHED) 
>>> + CMS_add1_signer(...) + CMS_final(...)’, but the problem was that I 
>>> systematically came across the ‘EVP_get_digestbyname’ function in the base 
>>> layers.
>>>
>>> If I understand correctly, the ‘EVP_get_digestbyname’ function is 
>>> incompatible with the OpenSSL 3+ provider system. I wanted to know if there 
>>> was a trick to make ‘EVP_get_digestbyname’ work or if there is a method 
>>> that I haven't seen to make a detached signature in DER format.
>>>
>> It's not "incompatible" with the provider system - but it is restricted. 
>> It only knows about certain built-in digests - but those digests can still 
>> come from a provider. Where in the CMS code are you hitting it?
>>
>> Matt
>>
>>  
>>
>>> Thank you for reading and your time.
>>>
>>> -- 
>>> You received this message because you are subscribed to the Google 
>>> Groups "openssl-users" group.
>>> To unsubscribe from this group and stop receiving emails from it, send 
>>> an email to [email protected]
>>> To view this discussion visit 
>>> https://groups.google.com/a/openssl.org/d/msgid/openssl-users/e3df5e46-f50d-4dcc-a852-38f2ca2014afn%40openssl.org 
>>> <https://groups.google.com/a/openssl.org/d/msgid/openssl-users/e3df5e46-f50d-4dcc-a852-38f2ca2014afn%40openssl.org?utm_medium=email&utm_source=footer>
>>> .
>>>
>>

-- 
You received this message because you are subscribed to the Google Groups "openssl-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openssl-users+unsubscribe-MCmKBN63+Bmbup2nOX2J7Q@public.gmane.org
To view this discussion visit https://groups.google.com/a/openssl.org/d/msgid/openssl-users/9aa99011-fd2f-4f37-851b-0df1a03a848bn%40openssl.org.