Which SM2 private key PEM file generated by openssl 3.x is correct?

[Eno-CN]

OpenSSL Version

OpenSSL 3.1.0 14 Mar 2023 (Library: OpenSSL 3.1.0 14 Mar 2023)

Description

I generated two different SM2 private key PEM files using two different methods.
When I use a tool to parse their ASN1 structures, I found that their OID identifiers are different, so I don't know which PEM file is generated correctly.

First Method

openssl genpkey -algorithm sm2 -text

PEM content:

-----BEGIN PRIVATE KEY-----
MIGIAgEAMBQGCCqBHM9VAYItBggqgRzPVQGCLQRtMGsCAQEEIPwnYlC0IWz13PYD
xisvfTJmzqJaKvVffqr0PGeQlEa9oUQDQgAEokJlXF9eQgJO6hvnvps0YQ8HPZot
6MCvh13WqDddbpZ0qey387GJ2PFBUghhucF9bu0bBZa52a6/Dbqc9gYfyA==
-----END PRIVATE KEY-----

ASN1 Structure:

Second Method

openssl genpkey -algorithm ec -pkeyopt ec_paramgen_curve:SM2 -text

PEM content:

-----BEGIN PRIVATE KEY-----
MIGHAgEAMBMGByqGSM49AgEGCCqBHM9VAYItBG0wawIBAQQgAzy83ltheFTNmJZX
bA9uR0d4hclB6BHdM3lnPglCLr6hRANCAARMGqLrNj095QswcJR3CEAZ2dvpAvyE
bWMBdn0woEtx9gjt+B6UperoEJfUVqm0g9VSdPrQbQEUcJzoq22b2eu9
-----END PRIVATE KEY-----

ASN1 Structure:

Additional Notes

However, I know that many applications in China support the second method of generating PEM files and cannot recognize the first PEM file.
The ASN1 structure of the SM2 private key PEM file generated by version 1.1.1 is the same as the one generated by the second method.

[t8m]

Maybe @InfoHunter could answer this?

[Eno-CN]

@InfoHunter Could you please help me answer a question if you have some time?

[zzl360]

I think #11328 could explained it.
In fact, the second method is the right way to satisfy chinese standard like GM/T 0015-2012 《Digital certificate format based on sm2 algorithm》。
this is one piece from that standard.

[Eno-CN]

I think #11328 could explained it.
In fact, the second method is the right way to satisfy chinese standard like GM/T 0015-2012 《Digital certificate format based on sm2 algorithm》。
this is one piece from that standard.

Ok, thank you so much.

[slontis]

I thought it was already established that 2 was correct. This will also interop with Bouncy Castle if we revert it back to this..

[lirc571]

In the same document (GM/T 0015-2012), table C.3 uses method 1's format, although it seems that most implementations use method 2.

[nhorman]

It seems like there is consensus that method 2 is correct, despite some documentation still using method 1.

As such, marking this as inactive, to be closed at the end of 3.4. Please comment further if there is more discussion needed