How to debug assembly? - linux

I have the following assembly file test which I want to debug,
How can I do that?
Note I am working with x86-64 and att syntax, plus I don't have access to c code.
I want to stop after each line and being able to see the registers in a table (I remember there is such an option).
I tried:
gdb test
r
but I get:
Starting program:
No executable file specified.
Use the "file" or "exec-file" command.

After running GDB on the executable1:
Use start or starti to set a breakpoint in main or _start respectively and run the program.
Or set breakpoints yourself with b 12 to set a breakpoint on source line 12 (if you built with enough debug info for this to work), or b *0x00401007 to set a breakpoint on an address you copy/pasted from disas output.
layout asm / layout reg puts GDB into text-UI mode with "windows" in your terminal for disassembly and registers. (This can be a bit flaky, you sometimes need control-L to redraw the screen, and sometimes GDB crashes when your process exits although I'm not sure if that's specifically from TUI.)
Otherwise without TUI mode, info reg and disas can be useful.
See the bottom of https://stackoverflow.com/tags/x86/info for more asm debugging tips.
Especially strace ./test is highly useful to see the system calls your program makes, decoded into C style. In toy programs you're playing with for your own experimentation, this basically works as an alternative to checking for error return values.
Footnote 1: You're not doing that part correctly:
No executable file specified.
That means no file called test existed in the directory where you ran gdb test.
You have to assemble + link test.S into an executable called test before you can run GDB on that file. If ls -l test shows it, then gdb test can debug it. (And ./test can run it.)
Often gcc -no-pie foo.S is a good choice to make debugging easier: addresses will be fixed at link time, so objdump -drwC -Mintel test output will match the addresses you see at run-time. And the addresses will be numerically smaller, so it's easier to visually spot a code (.text) address vs. .rodata (modern ld puts it in a separate page so it can avoid exec permission) vs. .data / .bss.
Either way, stack addresses are still easy to distinguish from code either way, 0x555... or 0x0000...XXXXXX is in the executable, 0x7fffff... is in the stack, other addresses from mmap are randomized. (But libc also gets mapped to a high address near the stack, with or without PIE.)
(Or if you're writing _start instead of main, gcc -nostdlib -static foo.S implies -no-pie)

Related

protecting stack memory of program

Lets assume that i wrote program in .c and that end user is starting .exe file. During program's execution there is variable called CHECK that gets dynamically assigned in middle of program's execution using some pseudo random algorithms. In one point, if the variable matches some criteria (lets say CHECK == 1580 or some static predifined number) the program does something on output. My question is, can a person that has controll over the system running this program modify memory in such way that he modifies address space of variable CHECK and match it to number '1580' before IF condition is set and trigger IF function even if the algorithm didnt set '1580' in the first place?
Yes, it's easy using a debugger, e.g. gdb. Set a breakpoint right before the if, run the program until the breakpoint triggers, set the variable to any desired value, remove the breakpoint, and continue. You could even have the debugger skip the condition check altogether, jumping directly into the if-block. You could also replace the check in the binary code by a nop. This is basically what "cracks" for pirating software do.
Without the source code and debugging symbols this becomes somewhat harder as you have to figure out the addresses, but it just delays the inevitable. With complete access to the computer, you can manipulate any program any way you want. Various protection schemes exist (mainly obfuscation), but they just make it harder, not impossible.
To further prove my point, here is a very quick example:
Given the following C code:
#include <stdlib.h>
#include <time.h>
#include <stdio.h>
int main () {
srand (time (NULL));
while (1) {
if (rand () == 1580) {
puts ("You got me!");
break;
}
}
}
Compile it with optimizations and without symbols to make it a little harder, assuming an x86_64 linux system:
gcc -O3 -flto -ffunction-sections -fdata-sections -Wl,--gc-sections -s test.c -o test
Ordinarily, this program would run for a few seconds before it quits. We want to make it quit immediately. Start it via the gdb debugger:
$ gdb ./test
(gdb) starti
Starting program: /tmp/test
Program stopped.
0x00007ffff7dd6090 in _start () from /lib64/ld-linux-x86-64.so.2
Obtain information about memory ranges. We are interested in the start address of the .text section:
(gdb) info files
Symbols from "/tmp/test".
Native process:
Using the running image of child process 12745.
While running this, GDB does not access memory from...
Local exec file:
`/tmp/test', file type elf64-x86-64.
Entry point: 0x555555554650
...
0x0000555555554610 - 0x00005555555547b2 is .text
...
So the actual code starts at 0x0000555555554610 in memory. Let's disassemble some of it:
(gdb) disas 0x0000555555554610,0x0000555555554700
Dump of assembler code from 0x555555554610 to 0x555555554700:
0x0000555555554610: xor %edi,%edi
0x0000555555554612: sub $0x8,%rsp
0x0000555555554616: callq 0x5555555545e0 <time#plt>
0x000055555555461b: mov %eax,%edi
0x000055555555461d: callq 0x5555555545d0 <srand#plt>
0x0000555555554622: nopl 0x0(%rax)
0x0000555555554626: nopw %cs:0x0(%rax,%rax,1)
0x0000555555554630: callq 0x5555555545f0 <rand#plt>
0x0000555555554635: cmp $0x62c,%eax
0x000055555555463a: jne 0x555555554630
0x000055555555463c: lea 0x17a(%rip),%rdi # 0x5555555547bd
0x0000555555554643: callq 0x5555555545c0 <puts#plt>
0x0000555555554648: xor %eax,%eax
0x000055555555464a: add $0x8,%rsp
0x000055555555464e: retq
...
That's the whole program. The cmp instruction is the interesting part; set a breakpoint there and let the program run:
(gdb) break *(0x0000555555554635)
Breakpoint 1 at 0x555555554635
(gdb) c
Continuing.
Breakpoint 1, 0x0000555555554635 in ?? ()
From the above assembly output you can see that 0x62c (i.e. 1580) is the magic number. Write it into the register, overwriting rand()s return value, and continue the program:
(gdb) set $eax = 1580
(gdb) c
Continuing.
You got me!
[Inferior 1 (process 12745) exited normally]
(gdb)
The program will immediately print the message and quit. Had we used some kind of password-entry function instead of just rand(), we could have done exactly the same thing to circumvent the password check. Instead of setting the value in the register, we could also have typed jump *0x000055555555463c to just jump into the if-block; that way, we don't even have to find the "magic" number.

Why is yasm generating incorrect debugging information?

I have an x86_64 assembly program I'm trying to debug on Linux, but when I try to use gdb, it skips around randomly and loops through the same couple instructions or repeats instructions. It also seems to loop through different instructions depending on where I set a breakpoint.
I tried researching this problem online, and I saw a number of people having this same issue with C++ when compilers were optimizing too aggressively and generating incorrect debugging information. I didn't see anything about assembly, but I believe yasm might be the problem here as well.
Here's my Makefile.
myprog : myprog.o
gcc -static -fdwarf2-cfi-asm myprog.o -o myprog
myprog.o : myprog.asm
yasm -f elf64 -g dwarf2 myprog.asm -o myprog.o
Note that I'm statically linking because I can't get dynamic linking to work. I might ask a separate question about that in the future.
And here's more or less what the gdb session looks like.
...
(gdb)# n
65 call findrepl
(gdb)# n
73 mov rdi, str3
(gdb)# n
75 call findrepl
(gdb)# n
75 call findrepl
(gdb)# n
65 call findrepl
...
Using s, si, or ni all do the same as n shown above. I haven't had this issue with my previous assembly programs. Sometimes gdb will say that I've executed something like xor eax, eax but then show the output from a call I made to printf.
I'm relatively new to assembly programming and gdb, so in the back of my mind I wonder if this is my fault. Is there a way to fix this? I would also like to know if there are any workarounds, as I can't seem to debug it without using gdb.
After much trial and error, I discovered the '-tui' option for gdb and the 'layout asm' command. There's probably a better way to do this, but here's basically what you type in to make it work:
[user#comp ~/prog/]$ gdb -tui myprog
(gdb)# layout asm
(gdb)# break main
(gdb)# run
And then you can go wild. By default, it's disassembled into AT&T notation. You can check which notation it's currently displaying with show disassembly-flavor. You can change it with set disassembly-flavor intel or set disassembly-flavor att.
For more information, run help tui, help layout, or help set disassembly-flavor in gdb.

How do you assemble, link and run a .s file in linux?

I'm getting a weird error message when trying to assemble and run a .s file using AT&T Intel Syntax. Not sure if I'm even using the correct architecture to begin with, or if I'm having syntax errors, if I'm not using the correct commands to assemble and link, etc. Completely lost and I do not know where to begin.
So basically, I have a file called yea.s , which contains some simple assembler instructions. I then try to compile it using the command as yea.s -o yea.o and then link is using ld yea.o -o yea. When running ld, I get this weird message:ld: warning: cannot find entry symbol _start; defaulting to 000000440000.
This is the program im trying to run, very simple and doesn't really do anything.
resMsg: .asciz "xxxxxxxx"
.text
.global main
main:
pushq $0
ret
I just cannot figure out what's going on. Obviously, this is for school homework. I'm not looking for the answer to the homework, obviously, but this is the starting point to where I can actually start the coding. And I just cant figure out how to simple run the program, which it doesn't say in the assignment. Anyway, thanks in advance guys!
Linux executables require an entry point to be specified. The entry point is the address of the first instruction to be executed in your program. If not specified otherwise, the link editor looks for a symbol named _start to use as an entry point. Your program does not contain such a symbol, thus the linker complains and picks the beginning of the .text section as the entry point. To fix this problem, rename main to _start.
Note further that unlike on DOS, there is nothing to return to from _start. So your attempt to return is going to cause a crash. Instead, call the system call sys_exit to exit the program:
mov $0, %edi # exit status
mov $60, %eax # system call number
syscall # perform exit call
Alternatively, if you want to use the C runtime environment and call functions from the C library, leave your program as is and instead assemble and link using the C compiler driver cc:
cc -o yea yea.s
If you do so, the C runtime environment provides the entry point for you and eventually tries to call a function main which is where your code comes in. This approach is required if you want to call functions from the C library. If you do it this way, make sure that main follows the SysV ABI (calling convention).
Note that even then your code is incorrect. The return value of a function is given in the eax (resp. rax) register and not pushed on the stack. To return zero from main, write
mov $0, %eax # exit status
ret # return from function
In all currently supported versions of Ubuntu open the terminal and type:
sudo apt install as31 nasm
as31: Intel 8031/8051 assembler
This is a fast, simple, easy to use Intel 8031/8051 assembler.
nasm: General-purpose x86 assembler
Netwide Assembler. NASM will currently output flat-form binary files, a.out, COFF and ELF Unix object files, and Microsoft 16-bit DOS and Win32 object files.
If you are using NASM in Ubuntu 18.04, the commands to compile and run an .asm file named example.asm are:
nasm -f elf64 example.asm # assemble the program
ld -s -o example example.o # link the object file nasm produced into an executable file
./example # example is an executable file

Is it possible to assemble and run raw CPU instructions using `as`?

There are a couple of related questions here.
Consider a program consisting only of the following two instructions
movq 1, %rax
cpuid
If I throw this into a file called Foo.asm, and run as Foo.asm, where as is the portable GNU assembler, I will get a file called a.out, of size 665 bytes on my system.
If I then chmod 700 a.out and try ./a.out, I will get an error saying cannot execute binary file.
Why is the file so large, if I am merely trying to translate two asm instructions into binary?
Why can the binary not be executed? I am providing valid instructions, so I would expect the CPU to be able to execute them.
How can I get exactly the binary opcodes for the asm instructions in my input file, instead of a bunch of extra stuff?
Once I have the answer to 3, how can I get my processor to execute them? (Assuming that I am not running privileged instructions.)
Why is the file so large, if I am merely trying to translate two asm instructions into binary?
Because the assembler creates a relocatable object file which includes additional information, like memory Sections and Symbol tables.
Why can the binary not be executed?
Because it is an (relocatable) object file, not a loadable file. You need to link it in order to make it executable so that it can be loaded by the operating system:
$ ld -o Foo a.out
You also need to give the linker a hint about where your program starts, by specifying the _start symbol.
But then, still, the Foo executable is larger than you might expect since it still contains additional information (e.g. the elf header) required by the operating system to actually launch the program.
Also, if you launch the executable now, it will result in a segmentation fault, since you are loading the contents of address 1, which is not mapped into your address space, into rax. Still, if you fix this, the program will run into undefined code at the end - you need to make sure to gracefully exit the program through a syscall.
A minimal running example (assumed x86_64 architecture) would look like
.globl _start
_start:
movq $1, %rax
cpuid
mov $60, %rax # System-call "sys_exit"
mov $0, %rdi # exit code 0
syscall
How can I get exactly the binary opcodes for the asm instructions in my input file, instead of a bunch of extra stuff?
You can use objcopy to generate a raw binary image from an object file:
$ objcopy -O binary a.out Foo.bin
Then, Foo.bin will only contain the instruction opcodes.
nasm has a -f bin option which creates a binary-only representation of your assembly code. I used this to implement a bare boot loader for VirtualBox (warning: undocumented, protoype only!) to directly launch binary code inside a VirtualBox image without operating system.
Once I have the answer to 3, how can I get my processor to execute them?
You will not be able to directly execute the raw binary file under Linux. You will need to write your own loader for that or not use an operating system at all. For an example, see my bare boot loader link above - this writes the opcodes into the boot loader of a VirtualBox disc image, so that the instructions are getting executed when launching the VirtualBox machine.
The old MS-DOS COM file format does not include a header. It really only contains the binary executable code. The code size can, however, not exceed 64kb. I don't know whether Linux can execute these.
You can write the opcodes into a file using a hexeditor. Then you just need to surround it with an elf header that Linux knows how to execute it.
Here's an example:
hexedit myfile.bin
Now just write your opcodes inside the file using the hexeditor.
After that you need to add the elf header. You could do this by hand and write the elf header into your .bin file, but that a bit tricky. Easiest method is to use a few commands (In this example for 64 bit).
objcopy --input-target=binary --output-target=elf64-x86-64 myfile.bin myfile.o
ld -o myfile myfile.o -T binary.ld
You will also need a linker script. I called this for example binary.ld.
And that are the contents of binary.ld:
ENTRY(_start);
SECTIONS
{
_start = 0x0;
}
Now you can execute your program: ./myfile
Perhaps there's something like exe2bin utility for the GNU tool set. I've used various versions of exe2bin with Microsoft tools, and the ARM toolkit has the ability to produce binaries, but I don't recall if it was directly from the linked output or something like exe2bin.

How to make gdb print symbols in shared libraries loaded with dlopen?

I want to debug a process running on Linux 2.6 using GDB. attach PID (where PID is the process ID), print main, print sin, print gzopen and print dlopen work (i.e. they find the respective symbols). But print myfoo doesn't work, where myfoo is a function loaded by the process from an .so file using dlopen. Here is what I get:
(gdb) print main
$3 = {int (int, char **)} 0x805ba90 <main>
(gdb) print sin
$4 = {<text variable, no debug info>} 0xb7701230 <sin>
(gdb) print gzopen
$5 = {<text variable, no debug info>} 0xb720df50 <gzopen>
(gdb) print dlopen
$6 = {<text variable, no debug info>} 0xb77248e0 <__dlopen_nocheck>
(gdb) print myfoo
No symbol "myfoo" in current context.
How do I get GDB to find myfoo?
The function myfoo does indeed exist, because in the program I managed to get its address using dlsym (after dlopen), and I managed to call it. Only after that I attached GDB to the process.
It turned out that there was a mydir/mylib.so: No such file or directory error message printed by the attach $PID command of GDB. Apparently GDB was started in the wrong directory. Doing the proper cd before starting GDB fixed the problem, and print myfoo started working.
I'd like to automate this: I want GDB figure out where my .so files (loaded with dlopen) are. An approximation I can think of is examining /proc/$PID/maps (on Linux), finding possible directories, and adding all of them to the GDB library search path before starting GDB. Extending LD_LIBRARY_PATH and doing a set solib-search-path /tmp/parent didn't work (ls -l /tmp/parent/mydir/myfoo.so does work), GDB still reported the No such file or directory. How do I tell GDB where to look for mydir/myfoo.so?
My other question is how do I get the list of possible directories? On Linux, /proc/$PID/maps contains them -- but what about other operating systems like FreeBSD and the Mac OS X?
"info target" command in gdb will show a list of all sections in all loaded shared objects (including dlopen()ed libraries). At least this works on Linux -- I don't know how it behaves on other operating systems.
I maintain a program that loads a shared library via dlopen() and have successfully accessed symbols in the shared library using GDB. This will only work, however, if the shared library has a symbol table.
It looks like there is no easy way to automate finding finding .so files in GDB.

Resources