Linux/Mac hashing SHA-1

Question

To do password exchange on the SQL server I need to use SHA-1 to generate a hash (no plain password exchange).

This is done differently on Linux/Mac so I abstracted it slightly.

ThorCryptWrapper.h

#ifndef THORS_ANVIL_MYSQL_DETAILS_THOR_CRYPTO_WRAPPER_H
#define THORS_ANVIL_MYSQL_DETAILS_THOR_CRYPTO_WRAPPER_H
#ifdef __APPLE__
#define COMMON_DIGEST_FOR_OPENSSL
#include <CommonCrypto/CommonDigest.h>
#define THOR_SHA1(data, len, dst) CC_SHA1(data, len, dst)
#else
#include <openssl/sha.h>
#define THOR_SHA1(data, len, dst) SHA1(data, len, dst)
#endif
namespace ThorsAnvil
{
 namespace MySQL
 {
typedef unsigned char ThorSHADigestStore[SHA_DIGEST_LENGTH];
inline void thorSHA1(ThorSHADigestStore& dest, ThorSHADigestStore& src)
{
 THOR_SHA1(src, SHA_DIGEST_LENGTH, dest);
}
inline void thorSHA1(ThorSHADigestStore& dest, std::string const& src)
{
 THOR_SHA1(reinterpret_cast<const unsigned char*>(&src[0]), src.length(), dest);
}
 }
}
#endif

• Jerry Coffin: Is there a really good reason you're writing code to use SHA-1? You do realize that collision attacks have been known for quite a while, and actually implemented to produce real collisions, right? This doesn't directly affect use for password hashing (AFAIK, no preimage attack has been published) but still points to a hash that's seriously compromised. If at all possible, I'd advise switching to something newer (e.g., SHA-256 or SHA-3).
• Loki Astari: @JerryCoffin: This is just implementing the MySQL "mysql_native_password" handshake requirements. They use SHA-1 as there specification. SHA1( password ) XOR SHA1( "20-bytes random data from server" <concat> SHA1( SHA1( password ) ) ). Documented here: dev.mysql.com/doc/internals/en/… You can see my use of it here: ThorsSQL Lib: Part 3: Layer 5 HandShake
• Jerry Coffin: For what it's worth: dev.mysql.com/doc/refman/5.6/en/…
• Loki Astari: @JerryCoffin: I know that's why in my code review I include: Currently this client does not support that and will throw an exception (A fun task for somebody that want to try :-) )

Answer 1

There's not a huge amount to review here.

---

It looks like the Apple code is intended to be a drop-in replacement for OpenSSL, so you could probably just rename to match, rather than creating a new name:

#ifdef __APPLE__
#define COMMON_DIGEST_FOR_OPENSSL
#include <CommonCrypto/CommonDigest.h>
#define SHA1 CC_SHA1
#else
#include <openssl/sha.h>
#endif

Whichever approach you take, don't forget to #undef the macro when you've finished using it (definitely before the end of the header), to avoid polluting the macro namespace for others.

---

I would prefer src.data() to &src[0] as the idiomatic way to access a string's characters as an array of char.

And I think I prefer sizeof src to SHA_DIGEST_LENGTH in the first overload, so that the connection is clear.

Is there any reason that src can't be a reference to const in both versions?

inline void thorSHA1(ThorSHADigestStore& dest, ThorSHADigestStore const& src)
{
 THOR_SHA1(src, sizeof src, dest);
}
inline void thorSHA1(ThorSHADigestStore& dest, std::string const& src)
{
 auto const src_bytes = reinterpret_cast<const unsigned char*>(src.data());
 THOR_SHA1(src_bytes, src.length(), dest);
}

---

I withhold comment on the suitability of SHA-1 for this purpose, as that appears to be something you're stuck with.