Hi,

I'm new to battle.net development, and I'm trying to implement the sha1 hashing for packet 0x51, however, because I already wrote sha1 for a torrent client I made, I assumed it would work for bnet, which now appears not to be the case.

Here is the code I'm using:
http://pastebin.org/192263

It includes the standard sha1 I implemented, and the sha1_bnet_encrypt() function which produces the hashes bnet expects. I would like to use the standard sha1 code if possible, so can any one tell me the offsets of the client's implementation, explain the differences, or both :). The code for the sha1_bnet_encrypt() function is a mess, so I'd rather not base the changes on that :(

Blizzard messed with it.  Check out JavaOp, iago's got some pretty easy to understand implementations of lots of different things Battle.net does weird: http://javaop.com/download.html

Okay, I'll look at that, thanks Rabbit.

Edit:

I just took a look at the javaop code. It's much closer to the standard sha1 code, just what I need, thanks again rabbit.

The differances between SHA1, X-SHA1 (Broken SHA1, Battle.net's Login implementation), L-SHA1 (Lockdown's Implementation of SHA1) are as follows, mainly they are endian issues. Standard SHA is Bigendian, Blizzard's is Littlendian.
[code]
//These two are the same, they just swapped the parameters
#define SHA1RoL(bits, word) \
    (((word) << (bits)) | ((word) >> (32-(bits))))
#define xSHA1RoL(word, bits) \
    (((word) << (bits)) | ((word) >> (32-(bits))))

//01 02 03 04 = 0x01020304
#define SHA1batoi(ba, i) \
  ((ba[i] << 24) | (ba[i+1] << 16) | (ba[i+2] << 8) | ba[i+3])

//01 02 03 04 = 0x04030201
#define xSHA1batoi(ba, i) \
  ((ba[i+3] << 24) | (ba[i+2] << 16) | (ba[i+1] << 8) | ba[i])
 
//0x01020304 = 01 02 03 04
#define SHA1itoba(a, ba, i) \
  (ba[i] = (uint8_t)(a >> 24)); (ba[i+1] = (uint8_t)(a >> 16)); (ba[i+2] = (uint8_t)(a >> 8)); (ba[i+3] = (uint8_t)a);

//0x01020304 = 04 03 02 01
#define xSHA1itoba(a, ba, i) \
  (ba[i+3] = (uint8_t)(a >> 24)); (ba[i+2] = (uint8_t)(a >> 16)); (ba[i+1] = (uint8_t)(a >> 8)); (ba[i] = (uint8_t)a);[/code]

So lets go down the list:
init() - Identical to standard
digest() - Lockdown and XSHA1 use littlendian when converting the intermediate hash to a byte array.
[code]  if(ctx->version != SHA1){
    for(i = 0; i < 5; i++)
      xSHA1itoba(ctx->intermediate_hash[i], digest, i * 4);
  }else{
    for(i = 0; i < 5; i++)
      SHA1itoba(ctx->intermediate_hash[i], digest, i * 4);
  }[/code]input() - Identical to standard
process_message_block() - The inners of the digest, Lockdown uses the littlendian byte to int array, and standard ROL. XSHA1 uses both littlendian and the reversed ROL.
[code]  if(ctx->version == xSHA1){
    for(t = 0; t < 16; t++)  W[t] = xSHA1batoi(ctx->message_block, t * 4);
    for(t = 16; t < 80; t++) W[t] = xSHA1RoL(1, W[t-3] ^ W[t-8] ^ W[t-14] ^ W[t-16]);
  }else if(ctx->version == lSHA1){
    for(t = 0; t < 16; t++)  W[t] = xSHA1batoi(ctx->message_block, t * 4);
    for(t = 16; t < 80; t++) W[t] = SHA1RoL(1, W[t-3] ^ W[t-8] ^ W[t-14] ^ W[t-16]);
  }else{
    for(t = 0; t < 16; t++)  W[t] = SHA1batoi(ctx->message_block, t * 4);
    for(t = 16; t < 80; t++) W[t] = SHA1RoL(1, W[t-3] ^ W[t-8] ^ W[t-14] ^ W[t-16]);
  }  [/code]
pad_message() - Then is the function that pads the buffer for the final digest. XSHA1 simply fills the remainder with nulls. Standard SHA, and Lockdown both pad with 0x80 and then nulls. The last 4 bytes are the length of all data pushed into the function. Lockdown uses Littlendian, Standard uses Bigendian.[code]void sha1_pad_message(sha1_context *ctx){
  if(ctx->version == xSHA1){
    while(ctx->message_block_index < 64)
      ctx->message_block[ctx->message_block_index++] = 0;
  }else{
    if (ctx->message_block_index > 55){
      ctx->message_block[ctx->message_block_index++] = 0x80;
 
      while(ctx->message_block_index < 64)
        ctx->message_block[ctx->message_block_index++] = 0;
   
      sha1_process_message_block(ctx);
    }else{
      ctx->message_block[ctx->message_block_index++] = 0x80;
    }
 
    while(ctx->message_block_index < 56)
      ctx->message_block[ctx->message_block_index++] = 0;

    if(ctx->version == lSHA1){
      xSHA1itoba(ctx->length_high, ctx->message_block, 60);
      xSHA1itoba(ctx->length_low,  ctx->message_block, 56);
    }else{
      SHA1itoba(ctx->length_high, ctx->message_block, 56);
      SHA1itoba(ctx->length_low,  ctx->message_block, 60);
    }
  }
  sha1_process_message_block(ctx);
}[/code]
Heres my full code:
sha1.h:[code]/*
*  sha1.h
*
*  Description:
*      This is the header file for code which implements the Secure
*      Hashing Algorithm 1 as defined in FIPS PUB 180-1 published
*      April 17, 1995.
*
*      Many of the variable names in this code, especially the
*      single character names, were used because those were the names
*      used in the publication.
*
*      Please read the file sha1.c for more information.
*
*  Notes:
*    I took this file from BNCSUtil's source, Why? Lazyness. I have
*    modified it a bit to support the 3 versions but most of its
*    from him. <3
*/

#ifndef _SHA1_H_
#define _SHA1_H_

#include "types.h"
#include "stdint.h"

#ifndef _SHA_enum_
#define _SHA_enum_
enum{
    shaSuccess = 0,
    shaNull,            /* Null pointer parameter */
    shaInputTooLong,    /* input data too long */
    shaStateError      /* called Input after Result */
};
#endif

#define sha1_hash_size 20

typedef enum{
  SHA1,
  xSHA1,
  lSHA1,
  wSHA1,
  MAX = 0xffffffff
}SHA1_t;
/*
*  This structure will hold context information for the SHA-1
*  hashing operation
*/

typedef struct sha1_context{
  uint32_t      intermediate_hash[sha1_hash_size / 4]; /* Message Digest                  */
  uint32_t      length_low;                            /* Message length in bits          */
  uint32_t      length_high;                          /* Message length in bits          */
  int_least16_t message_block_index;                  /* Index into message block array  */
  uint8_t      message_block[64];                    /* 512-bit message blocks          */
  uint8_t      computed;                              /* Is the digest computed?          */
  uint8_t      corrupted;                            /* Is the message digest corrupted? */
  SHA1_t        version;                              /* What version of SHA1 is this?    */
} sha1_context;

/* Function Prototypes */

int __stdcall sha1_reset(sha1_context *);
int __stdcall sha1_input(sha1_context *, const uint8_t *, uint32_t);
int __stdcall sha1_digest(sha1_context *, uint8_t *);
uint32_t __stdcall sha1_checksum(uint8_t *data, uint32_t length, uint32_t version);

#endif[/code]sha1.c:[code]/*
*  sha1.c
*
*  Description:
*    This file implements the Secure Hashing Algorithm 1 as
*    defined in FIPS PUB 180-1 published April 17, 1995.
*
*    The SHA-1, produces a 160-bit message digest for a given
*    data stream.  It should take about 2**n steps to find a
*    message with the same digest as a given message and
*    2**(n/2) to find any two messages with the same digest,
*    when n is the digest size in bits.  Therefore, this
*    algorithm can serve as a means of providing a
*    "fingerprint" for a message.
*
*  Portability Issues:
*    SHA-1 is defined in terms of 32-bit "words".  This code
*    uses <stdint.h> (included via "sha1.h" to define 32 and 8
*    bit unsigned integer types.  If your C compiler does not
*    support 32 bit unsigned integers, this code is not
*    appropriate.
*
*  Caveats:
*    SHA-1 is designed to work with messages less than 2^64 bits
*    long.  Although SHA-1 allows a message digest to be generated
*    for messages of any number of bits less than 2^64, this
*    implementation only works with messages with a length that is
*    a multiple of the size of an 8-bit character.
*
*  Notes:
*    I took this file from BNCSUtil's source, Why? Lazyness. I have
*    modified it a bit to support the 3 versions but most of its
*    from him. <3
*/

#include "sha1.h"

#define SHA1RoL(bits, word) \
    (((word) << (bits)) | ((word) >> (32-(bits))))
 
#define xSHA1RoL(word, bits) \
    (((word) << (bits)) | ((word) >> (32-(bits))))

#define SHA1batoi(ba, i) \
  ((ba[i] << 24) | (ba[i+1] << 16) | (ba[i+2] << 8) | ba[i+3])
 
#define xSHA1batoi(ba, i) \
  ((ba[i+3] << 24) | (ba[i+2] << 16) | (ba[i+1] << 8) | ba[i])
 
#define SHA1itoba(a, ba, i) \
  (ba[i] = (uint8_t)(a >> 24)); (ba[i+1] = (uint8_t)(a >> 16)); (ba[i+2] = (uint8_t)(a >> 8)); (ba[i+3] = (uint8_t)a);
 
#define xSHA1itoba(a, ba, i) \
  (ba[i+3] = (uint8_t)(a >> 24)); (ba[i+2] = (uint8_t)(a >> 16)); (ba[i+1] = (uint8_t)(a >> 8)); (ba[i] = (uint8_t)a);

/* Local Function Prototyptes */
void sha1_pad_message(sha1_context *);
void sha1_process_message_block(sha1_context *);

/*
*  SHA1Reset
*
*  Description:
*    This function will initialize the SHA1Context in preparation
*    for computing a new SHA1 message digest.
*
*  Parameters:
*    context: [in/out]
*      The context to reset.
*
*  Returns:
*    sha Error Code.
*
*/
int __stdcall sha1_reset(sha1_context *ctx){
  uint8_t x;
  if (!ctx)
    return shaNull;
 
  ctx->length_low          = 0;
  ctx->length_high          = 0;
  ctx->message_block_index  = 0;

  ctx->intermediate_hash[0] = 0x67452301;
  ctx->intermediate_hash[1] = 0xEFCDAB89;
  ctx->intermediate_hash[2] = 0x98BADCFE;
  ctx->intermediate_hash[3] = 0x10325476;
  ctx->intermediate_hash[4] = 0xC3D2E1F0;

  for(x = 0; x < 64; x++){
    ctx->message_block[x] = 0;
  }
  ctx->computed  = 0;
  ctx->corrupted = 0;

  return shaSuccess;
}

/*
*  SHA1Result
*
*  Description:
*    This function will return the 160-bit message digest into the
*    Message_Digest array  provided by the caller.
*    NOTE: The first octet of hash is stored in the 0th element,
*      the last octet of hash in the 19th element.
*
*  Parameters:
*    context: [in/out]
*      The context to use to calculate the SHA-1 hash.
*    Message_Digest: [out]
*      Where the digest is returned.
*
*  Returns:
*    sha Error Code.
*
*/
int __stdcall sha1_digest(sha1_context *ctx, uint8_t *digest){
  int i;

  if (!ctx || !digest)
    return shaNull;

  if (ctx->corrupted)
    return ctx->corrupted;

  if (!ctx->computed){
    sha1_pad_message(ctx);

    ctx->length_low  = 0;
    ctx->length_high = 0;
    ctx->computed    = 1;
  }

  if(ctx->version != SHA1){
    for(i = 0; i < 5; i++)
      xSHA1itoba(ctx->intermediate_hash[i], digest, i * 4);
  }else{
    for(i = 0; i < 5; i++)
      SHA1itoba(ctx->intermediate_hash[i], digest, i * 4);
  }

  return shaSuccess;
}
/*
*  SHA1Input
*
*  Description:
*    This function accepts an array of octets as the next portion
*    of the message.
*
*  Parameters:
*    context: [in/out]
*      The SHA context to update
*    message_array: [in]
*      An array of characters representing the next portion of
*      the message.
*    length: [in]
*      The length of the message in message_array
*
*  Returns:
*    sha Error Code.
*
*/
int __stdcall sha1_input(sha1_context *ctx, const uint8_t *data, uint32_t length){
  uint32_t x;
  if(!length)
    return shaSuccess;
 
  if(!ctx || !data)
      return shaNull;

  if(ctx->computed){
    ctx->corrupted = shaStateError;
    return shaStateError;
  }

  for(x = 0; x < length; x++){
    ctx->message_block[ctx->message_block_index++] = (data[x] & 0xFF);
    ctx->length_low += 8;

    if (ctx->length_low == 0){
      ctx->length_high++;
      if(ctx->length_high == 0){
        ctx->corrupted = shaInputTooLong;
        return shaInputTooLong;
      }
    }

    if (ctx->message_block_index == 64)
      sha1_process_message_block(ctx);
  }

  return shaSuccess;
}

/*
* SHA1Math
*
* Description:
*  This is simply so I can have a clean way of
* doing the Process in one loop insted of 4.
*/
uint32_t sha1_math(uint16_t t, uint32_t B, uint32_t C, uint32_t D){
  if(t < 20)      return ((B & C) | ((~B) & D));
  else if(t < 40) return (B ^ C ^ D);
  else if(t < 60) return ((B & C) | (B & D) | (C & D));
  else            return (B ^ C ^ D);
}
/*
*  SHA1ProcessMessageBlock
*
*  Description:
*    This function will process the next 512 bits of the message
*    stored in the Message_Block array.
*
*  Parameters:
*    None.
*
*  Returns:
*    Nothing.
*
*  Comments:
*    Many of the variable names in this code, especially the
*    single character names, were used because those were the
*    names used in the publication.
*
*
*/


void sha1_process_message_block(sha1_context *ctx){
  uint16_t t;            /* Loop counter        */
  uint32_t temp;          /* Temporary word value*/
  uint32_t W[80];        /* Word sequence      */
  uint32_t A, B, C, D, E; /* Word buffers        */
  const uint32_t K[] = {0x5A827999, 0x6ED9EBA1, 0x8F1BBCDC, 0xCA62C1D6};
 
  if(ctx->version == xSHA1){
    for(t = 0; t < 16; t++)  W[t] = xSHA1batoi(ctx->message_block, t * 4);
    for(t = 16; t < 80; t++) W[t] = xSHA1RoL(1, W[t-3] ^ W[t-8] ^ W[t-14] ^ W[t-16]);
  }else if(ctx->version == lSHA1){
    for(t = 0; t < 16; t++)  W[t] = xSHA1batoi(ctx->message_block, t * 4);
    for(t = 16; t < 80; t++) W[t] = SHA1RoL(1, W[t-3] ^ W[t-8] ^ W[t-14] ^ W[t-16]);
  }else{
    for(t = 0; t < 16; t++)  W[t] = SHA1batoi(ctx->message_block, t * 4);
    for(t = 16; t < 80; t++) W[t] = SHA1RoL(1, W[t-3] ^ W[t-8] ^ W[t-14] ^ W[t-16]);
  }
 
  A = ctx->intermediate_hash[0];
  B = ctx->intermediate_hash[1];
  C = ctx->intermediate_hash[2];
  D = ctx->intermediate_hash[3];
  E = ctx->intermediate_hash[4];
 
  for(t = 0; t < 80; t++){
    temp = SHA1RoL(5,A) + sha1_math(t, B, C, D) + E + W[t] + K[t / 20];
    E = D;
    D = C;
    C = SHA1RoL(30,B);
    B = A;
    A = temp;
  }
 
  ctx->intermediate_hash[0] += A;
  ctx->intermediate_hash[1] += B;
  ctx->intermediate_hash[2] += C;
  ctx->intermediate_hash[3] += D;
  ctx->intermediate_hash[4] += E;

  ctx->message_block_index = 0;
}


/*
*  SHA1PadMessage
*
*  Description:
*    According to the standard, the message must be padded to an even
*    512 bits.  The first padding bit must be a '1'.  The last 64
*    bits represent the length of the original message.  All bits in
*    between should be 0.  This function will pad the message
*    according to those rules by filling the Message_Block array
*    accordingly.  It will also call the ProcessMessageBlock function
*    provided appropriately.  When it returns, it can be assumed that
*    the message digest has been computed.
*
*  Parameters:
*    context: [in/out]
*      The context to pad
*    ProcessMessageBlock: [in]
*      The appropriate SHA*ProcessMessageBlock function
*  Returns:
*    Nothing.
*
*/

void sha1_pad_message(sha1_context *ctx){
  if(ctx->version == xSHA1){
    while(ctx->message_block_index < 64)
      ctx->message_block[ctx->message_block_index++] = 0;
  }else{
    if (ctx->message_block_index > 55){
      ctx->message_block[ctx->message_block_index++] = 0x80;
 
      while(ctx->message_block_index < 64)
        ctx->message_block[ctx->message_block_index++] = 0;
   
      sha1_process_message_block(ctx);
    }else{
      ctx->message_block[ctx->message_block_index++] = 0x80;
    }
 
    while(ctx->message_block_index < 56)
      ctx->message_block[ctx->message_block_index++] = 0;

    if(ctx->version == lSHA1){
      xSHA1itoba(ctx->length_high, ctx->message_block, 60);
      xSHA1itoba(ctx->length_low,  ctx->message_block, 56);
    }else{
      SHA1itoba(ctx->length_high, ctx->message_block, 56);
      SHA1itoba(ctx->length_low,  ctx->message_block, 60);
    }
  }
  sha1_process_message_block(ctx);
}



uint32_t __stdcall sha1_checksum(uint8_t *data, uint32_t length, uint32_t version){
  uint8_t digest[20];
  sha1_context ctx;
  ctx.version = version;
  sha1_reset(&ctx);
  sha1_input(&ctx, data, length);
  sha1_digest(&ctx, digest);
 
  return *((uint32_t*)&digest[0]) ^ *((uint32_t*)&digest[4]) ^
    *((uint32_t*)&digest[8]) ^ *((uint32_t*)&digest[12]) ^ *((uint32_t*)&digest[16]);
}[/code]

Firstly, thanks for the explanation and example code Hdx.

I tried modifying my code as per your instructions, however It still produces the wrong hashes. No doubt this is a failing on my behalf.

The modified code:
http://pastebin.org/195121

I do have some questions regarding your example, for instance, is the length of the message still encoded in big endian?

Edit:
I'm going to try running both the codes side by side and compare them.

Okay, I've got it working. It seems the problems were mainly due to endianess and inclusion of the message length which bnet does not encode.

The working code as it stands:
http://pastebin.org/195541

So thanks Hdx for the example code and exposition, and also to rabbit for the javaop link.