Introduction
This tutorial is more of a tip than a tutorial. It just explains how to calculate offsets for jumps and calls within the program you are patching.
General description
Nothing special much. It just calculates the GIMP offsets corresponding to Photoshop style angles and distances. Which might be very useful for drop shadow scripts and other scripts that use 3D effects. You simply enter the angle and distance and The GIMP will prompt you with two values, the X offset and Y offset, which are the mirrored offsets of where the light would be according to the angle and distance you entered. This means you will be able to use these values without any alteration other than possibly rounding them in for example the drop-shadow script. You'll find this new script in the Toolbox in the Xtns menu.
Precission notice
As with most computer programs, The GIMP doesn't support endless precission. Those precission errors can be disregarded in most cases as they are simply to small to be any serious problem. In the case of for example a 90 degrees angle however, where the X offset would be 0 under normal (read analog) circumstances, you'll get a number xe-y. Something like -6.12303e-17 for example. The e-17 means the puntuation mark will need to be moved to the left 17 positions, in other words the value should be divided by 100000000000000000, resulting in a value so close to 0, that 0 may be assumed.
As far as the script has been tested, this only occurs at multiples of 90 degrees angles. So beware when looking at resulting offset values to make sure there isn't an e-y at the end!
About the "Show Results" feature
Simply leave it on at all times. This feature has been included to make the script useable as a library function. If you uncheck this, the values will still be calculated, but no result will be shown on screen.
Note for developers
If you are a GIMP Action developer and want to be able to work with angles the way Photoshop does, instead of using the offset values which are more of a standard in The GIMP you should download and install this action. Simply calling the function (script-fu-degrees-to-offsets-calc angle distance) once installed will result in a list of offsets with the X offset first and Y offset second. Of course the requirement for the users using your resulting Action would be to have this script installed as well.
Types of Jumps/Calls
Here I will just describe the different types of jumps and calls which you will come across:
Short Jumps
Short jumps be they conditional or unconditional jumps are 2 bytes long (or 1 nibble if your Californian ;-). These are relative jumps taken from the first byte after the two bytes of the jump. Using short jumps you can jump a maximum of 127 bytes forward and 128 bytes backwards.
Long Jumps
Long jumps if they are relative are 6 bytes long for conditional jumps and are 5 bytes long for unconditional jumps. For conditional jumps 2 bytes are used to identify that it is a long jump and what type of jump (je, jg, jns etc) it is. The other 4 bytes are used to show how far away the target location is relative to the first byte after the jump. In an unconditional jump only 1 byte is used to identify it as a long unconditional jump and the other 4 are used to show it's target's relative position, as with the conditional jumps.
Calls
There are two different types of calls which we will use. The normal type of call works the same as the long jumps in that it is relative to it's current position. The other type gives a reference to a memory location, register or stack position which holds the memory location it will call. The position held by the later is direct e.g. the memory location referenced may contain 401036h which would be the exact position that you would call, not relative to the position of the call. The size of these types of calls depends on any calculations involved in the call i.e. you could do: 'call dword ptr [eax * edx + 2]'. Long jumps can also be made using this method, but I didn't say that earlier as to avoid repetition.
Tables
Here is a brief list of all the different types of jumps/calls and their appropriate op-codes. Where different jumps have the same Op-Codes I have grouped them:
Jump Description Short Op-Code Long Op-Code
call procedure call E8xxxxxxxx N/A
jmp u nconditional jump EBxx E9xxxxxxxx
ja/jnbe jump if above 77xx 0F87xxxxxxxx
jae/jnb/jnc jump if above or equal 73xx 0F83xxxxxxxx
jb/jc/jnae jump if below 72xx 0F82xxxxxxxx
jbe/jna jump if below or equal 76xx 0F86xxxxxxxx
jcxz/jecxz jump if cx/ecx equals zero E3xx N/A
je/jz jump if equal/zero 74xx 0F84xxxxxxxx
jne/jnz jump if not equal/zero 75xx 0F85xxxxxxxx
jg/jnle jump if greater 7Fxx 0F8Fxxxxxxxx
jge/jnl jump if greater or equal 7Dxx 0F8Dxxxxxxxx
jl/jnge jump if less 7Cxx 0F8Cxxxxxxxx
jle/jng jump if less or equal 7Exx 0F8Exxxxxxxx
jno jump if not overflow 71xx 0F81xxxxxxxx
jnp/jpo jump if no parity/parity odd 7Bxx 0F8Bxxxxxxxx
jns jump if not signed 79xx 0F89xxxxxxxx
jo jump if overflow 70xx 0F80xxxxxxxx
jp/jpe jump if parity/parity even 7Axx 0F8Axxxxxxxx
js jump if sign 78xx 0F88xxxxxxxx
Calculating Offsets (finding in the xx's in table)
You will need to be able to calculate offsets when you add jumps and make calls within and to the code you have added. If you choose to do this by hand instead of using a tool then here are the basics:
For jumps and calls further on in memory from your current position you take the address where you want to jump/call and subtract from it the memory location of the next instruction after your call/jump i.e.:
(target mem address) - (mem location of next instruction after call/jump)
Example
If we wanted to jump to 4020d0 and the next instruction *after* the jump is at location 401093 then we would use the following calculation:
4020d0 - 401093 = 103d
We then write the jump instruction in hex as e93d100000 where e9 is the hex op-code for a long relative jump and 3d100000 is the result of our calculation expanded to dword size and reversed.
For jumps and calls to locations *before* the current location in memory you take the address you want to call/jump to and subtract it from the memory location of the next instruction after your call/jump, then subtract 1 and finally perform a logical NOT on the result i.e.
NOT(mem address of next instruction - target mem address - 1)
Example
If we wanted to call location 401184 and the address of the next instruction after the call is 402190 then we do the following calculation:
NOT(402190 - 401184 - 1 ) = ffffeff4
We can then write our call instruction in hex as e8f4efffff where e8 is the hex op-code for relative call and f4efffff is the result of the calculation in reverse order.
If you want to practice with different examples then the best way to do this is to use a disassembler like WDASM which shows you the op-codes and try and work out the results yourself. Also as an end note you don't have to perform these calculations if you have enough room to make your jump or call instruction into an absolute jump call by doing the following as represented in assembler:
mov eax, 4020d0
call eax (or jmp eax)
Final Notes
Make life easier and use a program to do this ;-)
No comments:
Post a Comment