
                                         888    ,e,
         e88'888 888,8,  ,"Y88b  e88'888 888 ee  "  888 8e   e88 888
        d888  '8 888 "  "8" 888 d888  '8 888 P  888 888 88b d888 888
        Y888   , 888    ,ee 888 Y888   , 888 b  888 888 888 Y888 888
         "88,e8' 888    "88 888  "88,e8' 888 8b 888 888 888  "88 888
                                                              ,  88P
           Qapla's cracking tutorial, version 0.1 rel 970209  "8",P"

Introduction

     Welcome to my first attempt to write a Windows 95 cracking tutorial.

     This file is not meant as an introduction to either SoftIce, assembler
     or cracking in general. I will assume that you have installed SoftIce
     2.0 or 3.0 and that you are familiar with it. Some assembler and Win32
     API knowledge is also useful. If you are new to cracking, before
     continuing please read some of the files on cracking already available
     on the net, for example ED!SON's excellent tutorial. In his tutorial
     you will find an introduction to SoftIce, how to load exports and much
     more.

The program

     In this tutorial, I will use a great little program that you probably
     will find on the net by doing a simple search for it. The program is
     called StartClean, and the version I use is 1.2. The program scans the
     Windows 95 Start Menu and removes all shortcuts that don't point to
     anything. This is actually a very handy utility for those with a lot
     of software passing through their harddisks (like me), so this is one
     of the few little utilities I actually use. Another great thing about
     this program is that it is only 31kb, so it doesnt hog massive amounts
     of my harddrive. You *might* find this program attached to this
     tutorial.
     When you start the program it will fire up with a little nag-screen
     asking you to register it if you use it for more than 30 days. Even if
     we will defeat this protection several times in this file, I'm asking
     you that if you start using the program, please register it. The
     author deserves the money he is asking for it.

The extremly simplistic approach

     In this section I will use a method that works with this program, but
     it wont work with most other programs. I included it here to show you
     that there is no need to make anything more difficult then nessesary.
     (this is a good philosophy of life by the way :)

     Fire up the program, and press 'Register...'
     The program will show you a small dialog-box, asking you to enter your
     name and secret code. Now enter your name and any code. I entered
     "Qapla'97" and "115522". Press OK and it will tell you that the code
     was incorrect.

     Now comes the interesting part.
     In the explorer press the right mouse button on the file, and select
     Quick-View. A window will pop up with a lot of information about the
     file. The section we are interested in is the 'Import Table'. Scroll
     down until you reach this section.
     You will hopefully see something like this :

     Import Table

          COMCTL32.dll
          Ordinal  Function Name

          KERNEL32.dll
          Ordinal  Function Name
          026c     lstrcmpiA
          00d7     GetFileAttributesA
          026f     lstrcpyA
          0045     DeleteFileA
          0269     lstrcmpA
          01c1     RemoveDirectoryA
          .        .
          .        .
          .        .

     This section displays the API's the file uses. By setting a breakpoint
     on any of these you will be able to intercept the program when it uses
     them.
     Here comes the good part. The program somewhere in the code probably
     compares the code you entered with a pregenerated code, previously
     calculated from the name you entered.
     What does the 'lstrcmpA' function do? Lets look in the API-reference
     (the file I use is called Win32.hlp from the Win95-SDK, distributed
     with most real development environments, for example Borlands
     excellent Delphi 2.0)

     --- From Win32.hlp ---

     The lstrcmp function compares two character strings. The comparison is
     case sensitive.

       int lstrcmp(
           LPCTSTR  lpString1,     // address of first string
           LPCTSTR  lpString2      // address of second string
       );

       Parameters

       lpString1
         Points to the first null-terminated string to be compared.

       lpString2
         Points to the second null-terminated string to be compared.

       Return Value

       If the function succeeds and the string pointed to by lpString1
       is less than the string pointed to by lpString2, the return value
       is negative; if the string pointed to by lpString1 is greater than
       the string pointed to by lpString2, it is positive. If the strings
       are equal, the return value is zero.

     ---- End ---

     So, lets try setting a breakpoint on 'lstrcmpA'. Press ^D, and when
     the SoftIce screen appears type 'BPX lstrcmpA', now press ^D again and
     press OK once more. Blam, we were kicked back to SoftIce.
     ** Break due to KERNEL32!lstrcmp

     Now press F12 to return to the calling function, and you should see
     something like this:

                                  .
                                  .
                                  .
     0157:004011DD 50             PUSH EAX     <- push your code on the stack
     0157:004011DE 6830604000     PUSH 406030  <- push the right code on the stack
     0157:004011E3 FF1520924000   CALL [KERNEL32!lstrcmp] <- compare them
     0157:004011E9 85C0           TEST EAX, EAX
     0157:004011EB 0F8580000000   JNZ 00401271  <- check if they were the same
                                  .
                                  .

     At this point we have two options:

     a) Patch the JNZ to NOP's          This will make the program register
                                        with any code. This *may* introduce
                                        other problems, most noteably it
                                        might have a similar unpatched
                                        check in another part of the
                                        program that you won't notice.
     b) Find out the code it compared   This is a much better way of
     your code with                     working as you dont need to change
                                        the code and the serial you find
                                        will probably work with the next
                                        version of the software as well,
                                        the crack will probably stop
                                        working when you upgrade.

     Alternative (a) is left as an exercise to the reader :)

     Now type 'd 406030' <- this was the address it pushed on the stack,
     remember?

     The data-window will now display the correct code, in my case
     1398-13026-1211-249

     As i said in the beginning of this section, setting a breakpoint on
     string-compare API's will seldom work, as most programs use their own
     routine for doing this. The next section will present another, very
     similar approach to the same problem, but it will not rely on the same
     API.

The hmemcpy-bpm approach

     If you registered the program in the following section and wish to
     'unregister' it, so you can try this approach as well, you can start
     regedit, and delete the following key :
     HKEY_CURRENT_USER\SOFTWARE\Start Clean\Configuration

     Do a 'BC *' to clear all your existing breakpoints, and enter your
     name in the registration-box once again. (be sure to use an incorrect
     code, as we dont want to register it right now). Don't press OK yet.

     Now enter 'BPX hmemcpy', and press OK in the dialog. We will be back
     in SoftIce. Note that we are no longer in the flat addressing mode.
     This is protected mode 16bit code, ie in another context. We need to
     get back into the flat code before we can search all memory, but
     before we do that we will press ^D once again. The program just
     scanned our name this time, and we are just interested in is setting a
     breakpoint to trap access to the code we entered.

     We will shortly be back into softice again. Now press F12 a few times
     until we reach the 32bit code. You will notice this by looking at the
     addresses in the code window...

     0137:9EA6      <- this is a segmented 16-bit address.
     0157:004011B5  <- this is a 32-bit flat address.

     When you reach this code we can scan for the text we entered in the
     code-window. (you entered something unique didn't you, as we will be
     searching all physical memory, and a code like 0000 will probably be
     found in a lot of unrelated locations) Enter "s 0 l ffffffff
     'your_code_here'" and press enter.

     Now two things can happen. either it finds your code in a low address,
     (and this is what we are looking for), or it will find it somewhere
     around 0x80000000 (this is Windows internal memory-space, and not what
     we are looking for (Windows reserves the upper 2gig for internal use,
     and non ring 0 code will only have access to memory in the lower part
     of the address space))

     When you found what looks like the right place in memory, (I found it
     at 015f:0063f580), we will set a breakpoint for memory access there.

     Use 'BPM 0063f580' (or whatever address you found).

     Don't forget to 'BD hmemcpy' as well, as we will not be needing that
     breakpoint any more.

     Press ^D and you it will stop right in the function that compares the
     two strings.

     This method is usually much better than the previous, as it doesn't
     assume that the program uses any specific API's. It is usually safe to
     set a breakpoint on hmemcpy as almost all Win32-programs rely on this
     function to retrieve information from dialogs.

Other ways

     So, we have now defeated this program in two similar ways, and at this
     point I am starting to realize my bad choice of program as this little
     program doesn't contain any strange or non-standard things. It is
     rather unusually simplistic. If you feel like making a keymaker, which
     is the thing any *real* cracker would do, you can find the entrypoint
     to the code-generating routine just above the call to lstrcmp.

Setting breakpoints

     In ED!SON's tutorial, the author talks about the problems of setting
     breakpoints, especially when Norton Commander is active. When you try
     to do a BPX GetDlgItemTextA, you might get the 'No LDT' error. DOS
     windows, and especially Norton Commander hogs much of the CPU and if
     you are running them, there is a good chance you will end up in a VDM
     instead of the PROT32-mode you want to be in. SoftIce 3.0 seems to
     handle this much smarter, so if you are having problems try installing
     the latest version of the debugger. This is an issue of
     address-contexts and an extensive discussion on the topic can be found
     in the documentation for SoftIce 3.0. If you are trying to set a
     breakpoint in the code you are debugging and it doesnt work, try to
     break on a general API, and press F12 until you reach the context you
     are looking for, and then set the breakpoint.

Recommended reading

     The reason I wrote this tutorial is that during the last years, I have
     read quite a few text on cracking by different authors. I always
     wanted to make something similar to make a small contribution to this,
     and hopefully make someone reach a higher level of knowledge in
     cracking.

     I would like to recommend some of the great text on cracking already
     available on the net :

     ED!SON's Cracking Tutorial       This is a great file that contains
                                      an introduction to debugging,
                                      SoftIce and cracking. If you havn't
                                      read it yet, do so now. This file is
                                      *very* recommended for everyone.
     +ORC's Cracking Tutorials        These files are split up in lessons,
                                      each one talking about a different
                                      approach or side of cracking. Most
                                      of the lessons are very much worth
                                      reading, even if I dont agree with
                                      him in the frequent discussions
                                      about languages like Delphi or the
                                      world in general :) They might be a
                                      bit hard to find as he seems to be a
                                      bit reluctant about placing all of
                                      them on the net.

Thanks

     The author would like to thank the following persons for helping him
     with debugging the text, and verifying the wannabee-cracker-author's
     theory's...

     [prizna], odin- and kOUGER - thanks!

Contacting me

     You dont, but you *might* be able to find me on US-EFnet IRC. Check
     for the nick qapla, it might be me.

     thank you for reading this far, I hope you enjoyed it. (c)1997, Qapla'
