Re: [PATCH v2] CryptoPkg: Add new API to retrieve commonName of X.509 certificate

["Long, Qin" <qin.long@intel.com>]

From: Laszlo Ersek [mailto:lersek@redhat.com]
Sent: Thursday, September 21, 2017 12:38 AM
To: Long, Qin <qin.long@intel.com>; Ye, Ting <ting.ye@intel.com>; Zhang, Chao B <chao.b.zhang@intel.com>
Cc: edk2-devel@lists.01.org
Subject: Re: [PATCH v2] CryptoPkg: Add new API to retrieve commonName of X.509 certificate

Hello Qin,

On 09/20/17 18:05, Qin Long wrote:
> v2: Update function interface to return RETURN_STATUS to represent
>     different error cases.
>
> Add one new API (X509GetCommonName()) to retrieve the subject commonName
> string from one X.509 certificate.
>
> Cc: Laszlo Ersek <lersek@redhat.com<mailto:lersek@redhat.com>>
> Cc: Ting Ye <ting.ye@intel.com<mailto:ting.ye@intel.com>>
> Cc: Chao Zhang <chao.b.zhang@intel.com<mailto:chao.b.zhang@intel.com>>
> Contributed-under: TianoCore Contribution Agreement 1.0
> Signed-off-by: Qin Long <qin.long@intel.com<mailto:qin.long@intel.com>>
> ---
>  CryptoPkg/Application/Cryptest/RsaVerify2.c        |  32 +++++--
>  CryptoPkg/Include/Library/BaseCryptLib.h           |  34 +++++++
>  CryptoPkg/Library/BaseCryptLib/Pk/CryptX509.c      | 106 +++++++++++++++++++++
>  CryptoPkg/Library/BaseCryptLib/Pk/CryptX509Null.c  |  32 +++++++
>  .../Pk/CryptX509Null.c                             |  34 ++++++-
>  5 files changed, 230 insertions(+), 8 deletions(-)
>
> diff --git a/CryptoPkg/Include/Library/BaseCryptLib.h b/CryptoPkg/Include/Library/BaseCryptLib.h
> index 9c5ffcd9cf..48e9531758 100644
> --- a/CryptoPkg/Include/Library/BaseCryptLib.h
> +++ b/CryptoPkg/Include/Library/BaseCryptLib.h
> @@ -2171,6 +2171,40 @@ X509GetSubjectName (
>    IN OUT  UINTN        *SubjectSize
>    );
>
> +/**
> +  Retrieve the common name (CN) string from one X.509 certificate.
> +
> +  @param[in]      Cert             Pointer to the DER-encoded X509 certificate.
> +  @param[in]      CertSize         Size of the X509 certificate in bytes.
> +  @param[out]     CommonName       Buffer to contain the retrieved certificate common
> +                                   name string. At most CommonNameSize bytes will be
> +                                   written and the string will be null terminated. May be
> +                                   NULL in order to determine the size buffer needed.
> +  @param[in,out]  CommonNameSize   The size in bytes of the CommonName buffer on input,
> +                                   and the size of buffer returned CommonName on output.
> +                                   If CommonName is NULL then the amount of space needed
> +                                   in buffer (including the final null) is returned.
> +
> +  @retval RETURN_SUCCESS           The certificate CommonName retrieved successfully.
> +  @retval RETURN_INVALID_PARAMETER If Cert is NULL.
> +                                   If CommonNameSize is NULL.
> +                                   If Certificate is invalid.
> +  @retval RETURN_NOT_FOUND         If no CommonName entry exists.
> +  @retval RETURN_BUFFER_TOO_SMALL  If the CommonName is NULL. The required buffer size
> +                                   (including the final null) is returned in the
> +                                   CommonNameSize parameter.
> +  @retval RETURN_UNSUPPORTED       The operation is not supported.
> +
> +**/
> +RETURN_STATUS
> +EFIAPI
> +X509GetCommonName (
> +  IN      CONST UINT8  *Cert,
> +  IN      UINTN        CertSize,
> +  OUT     CHAR8        *CommonName,
> +  IN OUT  UINTN        *CommonNameSize
> +  );
> +
>  /**
>    Verify one X509 certificate was issued by the trusted CA.
>

I think the RETURN_BUFFER_TOO_SMALL description is incorrect -- it
shouldn't only cover the (CommonName == NULL) case, but any other case
when *CommonNameSize is not large enough, for formatting the full CN,
plus the terminating '\0'.

Relatedly, the output value of *CommonNameSize should always be the
number of bytes required to format the NUL-terminated common name,
regardless if there is enough room or not. The return status will tell
the caller:
- if the return status is BUFFER_TOO_SMALL, then a larger buffer is
needed -- how large is explained by *CommonNameSize
- if the return status is SUCCESS, then the buffer was large enough, and
*CommonNameSize bytes have been used from it.

[qlong] good catch.
The current implementation is based on OpenSSL X509_NAME_get_text_by_OBJ
API, and we can only get the real written data size or required size (by passing
NULL CommonName) with this interface.
I didn’t want to introduce additional handling (e.g. extra ASN1_STRING parsing) in
this API. For fixed CommonNameSize buffer, it’s acceptable to receive the truncated
string (e.g. for display purpose) depending on the API usage (and the CN string should
be less than 64 char per the standard).

Additional question: does the code handle the case when *CommonNameSize
is zero, on input? (I.e., when there isn't even room for storing a '\0'
character.)

[qlong] It’s one missed case. And more interesting is this may expose one OpenSSL issue.
              Let me double-check this. ☺

Thanks,
Laszlo