Get machine code of the proccess by PID without attaching a debugger - linux

I want to get a machine code of the running proccess by his PID for analysing malicious instructions, by using heuristic methods of data analysing.
All I need to know is list of current machine instructions and values of registers (EIP, EAX, EBX...).
I can use gdb for reach this goal gdb output, but is take a several problems:
I don't know how interact with gdb from my application;
malicious code can use some technics of debugger detection like this: http://www.ouah.org/linux-anti-debugging.txt
https://www.youtube.com/watch?v=UTVp4jpJoyc&list=LLw7XNcx80oj63tRYAg7hrsA
for windows;
Getting info from console output makes work of my application slower.
Is are any way to get this information by PID in Linux? Or maybe Windows?

you may have a look to gcore:
$ gcore
usage: gcore [-o filename] pid
so you can dump process core using its pid:
$ gcore 792
warning: Could not load vsyscall page because no executable was specified
0x00007f5f73998410 in ?? ()
Saved corefile core.792
and then open it in gdb:
$ gdb -c core.792
GNU gdb (GDB) Fedora 8.0.1-30.fc26
Copyright (C) 2017 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
[...]
[New LWP 792]
Missing separate debuginfo for the main executable file
Try: dnf --enablerepo='*debug*' install /usr/lib/debug/.build-id/09/b9d38bb6291b6282de4a2692e45448828d50da
Core was generated by `./a.out'.
#0 0x00007f5f73998410 in ?? ()
(gdb) info registers
rax 0xfffffffffffffe00 -512
rbx 0x0 0
rcx 0x7f5f73998410 140047938061328
rdx 0x1 1
rsi 0x7ffd30683d73 140725415591283
rdi 0x3 3
rbp 0x7ffd30683d90 0x7ffd30683d90
rsp 0x7ffd30683d68 0x7ffd30683d68
r8 0x1d 29
r9 0x0 0
r10 0x3 3
r11 0x246 582
r12 0x4006d0 4196048
r13 0x7ffd30683e70 140725415591536
r14 0x0 0
r15 0x0 0
rip 0x7f5f73998410 0x7f5f73998410
eflags 0x246 [ PF ZF IF ]
cs 0x33 51
ss 0x2b 43
ds 0x0 0
es 0x0 0
fs 0x0 0
gs 0x0 0
or even using the binary image from /proc to get some symbols:
gdb -c core.792 /proc/792/exe
You may know that you can pass scripts to gdb, this can ease not having to interact with it from your binary (man gdb for more details).
if you don't want to use gdb directly you may try using ptrace() directly, but it is for sure more work.
For the anti debugging technics, well... they work and there is no easy way to handle them directly as far as I know, each one may be worked arounded manually, (patching binary, disassembling from unaligned addresses manually by setting then in objdump, etc...)
I'm not an expert of the domain, I hope this will help you a bit.

Related

"Whirlwind Tutorial on Teensy ELF Executables" -- why is the output of ld 10X bigger, 20 years later? [duplicate]

For a university course, I like to compare code-sizes of functionally similar programs if written and compiled using gcc/clang versus assembly. In the process of re-evaluating how to further shrink the size of some executables, I couldn't trust my eyes when the very same assembly code I assembled/linked 2 years ago now has grown >10x in size after building it again (which true for multiple programs, not only helloworld):
$ make
as -32 -o helloworld-asm-2020.o helloworld-asm-2020.s
ld -melf_i386 -o helloworld-asm-2020 helloworld-asm-2020.o
$ ls -l
-rwxr-xr-x 1 xxx users 708 Jul 18 2018 helloworld-asm-2018*
-rwxr-xr-x 1 xxx users 8704 Nov 25 15:00 helloworld-asm-2020*
-rwxr-xr-x 1 xxx users 4724 Nov 25 15:00 helloworld-asm-2020-n*
-rwxr-xr-x 1 xxx users 4228 Nov 25 15:00 helloworld-asm-2020-n-sstripped*
-rwxr-xr-x 1 xxx users 604 Nov 25 15:00 helloworld-asm-2020.o*
-rw-r--r-- 1 xxx users 498 Nov 25 14:44 helloworld-asm-2020.s
The assembly code is:
.code32
.section .data
msg: .ascii "Hello, world!\n"
len = . - msg
.section .text
.globl _start
_start:
movl $len, %edx # EDX = message length
movl $msg, %ecx # ECX = address of message
movl $1, %ebx # EBX = file descriptor (1 = stdout)
movl $4, %eax # EAX = syscall number (4 = write)
int $0x80 # call kernel by interrupt
# and exit
movl $0, %ebx # return code is zero
movl $1, %eax # exit syscall number (1 = exit)
int $0x80 # call kernel again
The same hello world program, compiled using GNU as and GNU ld (always using 32-bit assembly) was 708 bytes then, and has grown to 8.5K now. Even when telling the linker to turn off page alignment (ld -n), it still has almost 4.2K. stripping/sstripping doesn't pay off either.
readelf tells me that the start of section headers is much later in the code (byte 468 vs 8464), but I have no idea why. It's running on the same arch system as in 2018, the Makefile is the same and I'm not linking against any libraries (especially not libc). I guess something regarding ld has changed due to the fact that the object file is still quite small, but what and why?
Disclaimer: I'm building 32-bit executables on an x86-64 machine.
Edit: I'm using GNU binutils (as & ld) version 2.35.1 Here is a base64-encoded archive which includes the source and both executables (small old one, large new one) :
cat << EOF | base64 -d | tar xj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EOF
Update:
When using ld.gold instead of ld.bfd (to which /usr/bin/ld is symlinked to by default), the executable size becomes as small as expected:
$ cat Makefile
TARGET=helloworld
all:
as -32 -o ${TARGET}-asm.o ${TARGET}-asm.s
ld.bfd -melf_i386 -o ${TARGET}-asm-bfd ${TARGET}-asm.o
ld.gold -melf_i386 -o ${TARGET}-asm-gold ${TARGET}-asm.o
rm ${TARGET}-asm.o
$ make -q
$ ls -l
total 68
-rw-r--r-- 1 eso eso 200 Dec 1 13:57 Makefile
-rwxrwxr-x 1 eso eso 8700 Dec 1 13:57 helloworld-asm-bfd
-rwxrwxr-x 1 eso eso 732 Dec 1 13:57 helloworld-asm-gold
-rw-r--r-- 1 eso eso 498 Dec 1 13:44 helloworld-asm.s
Maybe I just used gold previously without being aware.
It's not 10x in general, it's page-alignment of a couple sections as Jester says, per changes to ld's default linker script for security reasons:
First change: Making sure data from .data isn't present in any of the mapping of .text, so none of that static data is available for ROP / Spectre gadgets in an executable page. (In older ld, that meant the program-headers mapped the same disk-block twice, also into a RW-without-exec segment for the actual .data section. The executable mapping was still read-only.)
More recent change: Separate .rodata from .text into separate segments, again so static data isn't mapped into an executable page. Previously, const char code[]= {...} could be cast to a function pointer and called, without needing mprotect or gcc -z execstack or other tricks, if you wanted to test shellcode that way. (A separate Linux kernel change made -z execstack only apply to the actual stack, not READ_IMPLIES_EXEC.)
See Why an ELF executable could have 4 LOAD segments? for this history, including the strange fact that .rodata is in a separate segment from the read-only mapping for access to the ELF metadata.
That extra space is just 00 padding and will compress well in a .tar.gz or whatever.
So it has a worst-case upper bound of about 2x 4k extra pages of padding, and tiny executables are close to that worst case.
gcc -Wl,--nmagic will turn off page-alignment of sections if you want that for some reason. (see the ld(1) man page) I don't know why that doesn't pack everything down to the old size. Perhaps checking the default linker script would shed some light, but it's pretty long. Run ld --verbose to see it.
stripping won't help for padding that's part of a section; I think it can only remove whole sections.
ld -z noseparate-code uses the old layout, only 2 total segments to cover the .text and .rodata sections, and the .data and .bss sections. (And the ELF metadata that dynamic linking wants access to.)
Related:
Linking with gcc instead of ld
This question is about ld, but note that if you're using gcc -nostdlib, that used to also default to making a static executable. But modern Linux distros config GCC with -pie as the default, and GCC won't make a static-pie by default even if there aren't any shared libraries being linked. Unlike with -no-pie mode where it will simply make a static executable in that case. (A static-pie still needs startup code to apply relocations for any absolute addresses.)
So the equivalent of ld directly is gcc -nostdlib -static (which implies -no-pie). Or gcc -nostdlib -no-pie should let it default to -static when there are no shared libs being linked. You can combine this with -Wl,--nmagic and/or -Wl,-z -Wl,noseparate-code.
Also:
A Whirlwind Tutorial on Creating Really Teensy ELF Executables for Linux - eventually making a 45 byte executable, with the machine code for an _exit syscall stuffed into the ELF program header itself.
FASM can make quite small executables, using its mode where it outputs a static executable (not object file) directly with no ELF section metadata, just program headers. (It's a pain to debug with GDB or disassemble with objdump; most tools assume there will be section headers, even though they're not needed to run static executables.)
What is a reasonable minimum number of assembly instructions for a small C program including setup?
What's the difference between "statically linked" and "not a dynamic executable" from Linux ldd? (static vs. static-pie vs. (dynamic) PIE that happens to have no shared libraries.)

Minimal executable size now 10x larger after linking than 2 years ago, for tiny programs?

For a university course, I like to compare code-sizes of functionally similar programs if written and compiled using gcc/clang versus assembly. In the process of re-evaluating how to further shrink the size of some executables, I couldn't trust my eyes when the very same assembly code I assembled/linked 2 years ago now has grown >10x in size after building it again (which true for multiple programs, not only helloworld):
$ make
as -32 -o helloworld-asm-2020.o helloworld-asm-2020.s
ld -melf_i386 -o helloworld-asm-2020 helloworld-asm-2020.o
$ ls -l
-rwxr-xr-x 1 xxx users 708 Jul 18 2018 helloworld-asm-2018*
-rwxr-xr-x 1 xxx users 8704 Nov 25 15:00 helloworld-asm-2020*
-rwxr-xr-x 1 xxx users 4724 Nov 25 15:00 helloworld-asm-2020-n*
-rwxr-xr-x 1 xxx users 4228 Nov 25 15:00 helloworld-asm-2020-n-sstripped*
-rwxr-xr-x 1 xxx users 604 Nov 25 15:00 helloworld-asm-2020.o*
-rw-r--r-- 1 xxx users 498 Nov 25 14:44 helloworld-asm-2020.s
The assembly code is:
.code32
.section .data
msg: .ascii "Hello, world!\n"
len = . - msg
.section .text
.globl _start
_start:
movl $len, %edx # EDX = message length
movl $msg, %ecx # ECX = address of message
movl $1, %ebx # EBX = file descriptor (1 = stdout)
movl $4, %eax # EAX = syscall number (4 = write)
int $0x80 # call kernel by interrupt
# and exit
movl $0, %ebx # return code is zero
movl $1, %eax # exit syscall number (1 = exit)
int $0x80 # call kernel again
The same hello world program, compiled using GNU as and GNU ld (always using 32-bit assembly) was 708 bytes then, and has grown to 8.5K now. Even when telling the linker to turn off page alignment (ld -n), it still has almost 4.2K. stripping/sstripping doesn't pay off either.
readelf tells me that the start of section headers is much later in the code (byte 468 vs 8464), but I have no idea why. It's running on the same arch system as in 2018, the Makefile is the same and I'm not linking against any libraries (especially not libc). I guess something regarding ld has changed due to the fact that the object file is still quite small, but what and why?
Disclaimer: I'm building 32-bit executables on an x86-64 machine.
Edit: I'm using GNU binutils (as & ld) version 2.35.1 Here is a base64-encoded archive which includes the source and both executables (small old one, large new one) :
cat << EOF | base64 -d | tar xj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EOF
Update:
When using ld.gold instead of ld.bfd (to which /usr/bin/ld is symlinked to by default), the executable size becomes as small as expected:
$ cat Makefile
TARGET=helloworld
all:
as -32 -o ${TARGET}-asm.o ${TARGET}-asm.s
ld.bfd -melf_i386 -o ${TARGET}-asm-bfd ${TARGET}-asm.o
ld.gold -melf_i386 -o ${TARGET}-asm-gold ${TARGET}-asm.o
rm ${TARGET}-asm.o
$ make -q
$ ls -l
total 68
-rw-r--r-- 1 eso eso 200 Dec 1 13:57 Makefile
-rwxrwxr-x 1 eso eso 8700 Dec 1 13:57 helloworld-asm-bfd
-rwxrwxr-x 1 eso eso 732 Dec 1 13:57 helloworld-asm-gold
-rw-r--r-- 1 eso eso 498 Dec 1 13:44 helloworld-asm.s
Maybe I just used gold previously without being aware.
It's not 10x in general, it's page-alignment of a couple sections as Jester says, per changes to ld's default linker script for security reasons:
First change: Making sure data from .data isn't present in any of the mapping of .text, so none of that static data is available for ROP / Spectre gadgets in an executable page. (In older ld, that meant the program-headers mapped the same disk-block twice, also into a RW-without-exec segment for the actual .data section. The executable mapping was still read-only.)
More recent change: Separate .rodata from .text into separate segments, again so static data isn't mapped into an executable page. Previously, const char code[]= {...} could be cast to a function pointer and called, without needing mprotect or gcc -z execstack or other tricks, if you wanted to test shellcode that way. (A separate Linux kernel change made -z execstack only apply to the actual stack, not READ_IMPLIES_EXEC.)
See Why an ELF executable could have 4 LOAD segments? for this history, including the strange fact that .rodata is in a separate segment from the read-only mapping for access to the ELF metadata.
That extra space is just 00 padding and will compress well in a .tar.gz or whatever.
So it has a worst-case upper bound of about 2x 4k extra pages of padding, and tiny executables are close to that worst case.
gcc -Wl,--nmagic will turn off page-alignment of sections if you want that for some reason. (see the ld(1) man page) I don't know why that doesn't pack everything down to the old size. Perhaps checking the default linker script would shed some light, but it's pretty long. Run ld --verbose to see it.
stripping won't help for padding that's part of a section; I think it can only remove whole sections.
ld -z noseparate-code uses the old layout, only 2 total segments to cover the .text and .rodata sections, and the .data and .bss sections. (And the ELF metadata that dynamic linking wants access to.)
Related:
Linking with gcc instead of ld
This question is about ld, but note that if you're using gcc -nostdlib, that used to also default to making a static executable. But modern Linux distros config GCC with -pie as the default, and GCC won't make a static-pie by default even if there aren't any shared libraries being linked. Unlike with -no-pie mode where it will simply make a static executable in that case. (A static-pie still needs startup code to apply relocations for any absolute addresses.)
So the equivalent of ld directly is gcc -nostdlib -static (which implies -no-pie). Or gcc -nostdlib -no-pie should let it default to -static when there are no shared libs being linked. You can combine this with -Wl,--nmagic and/or -Wl,-z -Wl,noseparate-code.
Also:
A Whirlwind Tutorial on Creating Really Teensy ELF Executables for Linux - eventually making a 45 byte executable, with the machine code for an _exit syscall stuffed into the ELF program header itself.
FASM can make quite small executables, using its mode where it outputs a static executable (not object file) directly with no ELF section metadata, just program headers. (It's a pain to debug with GDB or disassemble with objdump; most tools assume there will be section headers, even though they're not needed to run static executables.)
What is a reasonable minimum number of assembly instructions for a small C program including setup?
What's the difference between "statically linked" and "not a dynamic executable" from Linux ldd? (static vs. static-pie vs. (dynamic) PIE that happens to have no shared libraries.)

How can the intel system instructions be run on linux?

Using 7th gen i5 processor, nasm, ld.
Written using general purpose registers and system calls, the program runs on Ubuntu. An example is a simple hello world program.
global _start
section .text
_start:
mov eax, 0x4 ; write(int fd, char *buf, int len)
mov ebx, 0x1 ; fd
mov ecx, message ; *buf
mov edx, message_length ; len
int 0x80 ; syscall
mov eax, 0x1 ; exit(int status)
mov ebx, 0x0 ; status
int 0x80 ; syscall
section .data
message db "hello world", 0xA
message_length equ $-message
nasm -f elf64 -o hello_world.o hello_world.s
ld hello_world.o -o hello_world
./hello_world
output: hello_world
However, the program written using the intel system instructions does not work.
global _start
section .text
_start:
CLI
HLT
nasm -f elf64 halt.s -o halt.o
ld halt.o -o halt
./halt
output: Segmentation fault (core dumped)
What prevents this code from compiling this way and running?
How can this code be compiled and run?
You have two questions:
What prevents this code from compiling this way and running?
Nothing prevents it from assembling, because it is valid code.
The reason that it does not run/throws a segmentation fault is, that you are attempting to run instructions that require certain privileges in user mode.
Have a look at HLT which description states
The HLT instruction is a privileged instruction.
Also look at CLI/STI, whose situation is more complicated (which is explained in the below article and the comments below), but is also mostly useful for kernel mode code. Wikipedia says
In all three cases, only privileged applications (usually the OS kernel) may modify IF[Interrupt flag]. [...] CLI and STI are [also] privileged instructions, which trigger a general protection fault if an unprivileged application attempts to execute it, [...]
That should answer your question regarding the difference between these two instructions and why HLT surely generates a GPF (General Protection Fault) executed in User Mode.
How can this code be compiled and run?
The only way to run this code is in privileged mode. So on Windows or Linux you'd have to code a kernel driver or code your own OS to make sensible use of these instructions.
The above only applies to Protected Mode or Long Mode code.
(Real mode code may always modify IF, the Interrupt flag)

Why does strace report my x64 FASM program runs in 32-bit mode? [duplicate]

This question already has answers here:
What is better "int 0x80" or "syscall" in 32-bit code on Linux?
(3 answers)
Closed 3 years ago.
I have format ELF64 executable 3 at top of my source file.
I compiled my program using fasm main.asm
Output:
flat assembler version 1.73.13 (16384 kilobytes memory, x64)
3 passes, 319 bytes.
Then I tried to run it using strace ./main, because it didn't work as expected and in output there is strace: [ Process PID=3012310 runs in 32 bit mode. ].
file main: main: ELF 64-bit LSB executable, x86-64, version 1 (GNU/Linux), statically linked, no section header
uname -m: x86_64
Use syscall instead of int 0x80
strace is wrong, your process isn't actually running in 32-bit mode, just using the 32-bit int 0x80 system call ABI.
You can check with gdb ./main and use starti. info regs will show that the register state is 64-bit, including 16x 64-bit registers, not 8x 32-bit registers. Or more simply, layout reg.
I see the same strace bug(?) when building a program with NASM that uses the 32-bit int 0x80 ABI in 64-bit mode to make an exit system call.
I added a delay loop before the first system call and I see strace doesn't print out the bitness of the target process until it makes a system call. So apparently strace infers that from whether it uses the 64-bit syscall ABI or the 32-bit int 0x80 / sysenter ABI!
What happens if you use the 32-bit int 0x80 Linux ABI in 64-bit code?
Perhaps this is related to strace trying to figure out how to decode system calls: The Linux ptrace API that strace uses doesn't have a simple reliable mechanism to tell which system call ABI a process invoked. https://superuser.com/questions/834122/how-to-distinguish-syscall-from-int-80h-when-using-ptrace
A 64-bit process that uses 32-bit system calls used to just get decoded according to 64-bit call numbers. But now it seems modern strace checks:
I used eax=1 / syscall to invoke write, and eax=1 / int 0x80 to invoke exit, and strace decoded them both correctly
execve("./nasm-test", ["./nasm-test"], 0x7ffdb8da5890 /* 52 vars */) = 0
write(0, NULL, 0) = 0
strace: [ Process PID=5219 runs in 32 bit mode. ]
exit(0) = ?
+++ exited with 0 +++
This is with strace 5.3 on Linux 5.3.1-arch1-1-ARCH.

How to single step ARM assembly in GDB on QEMU?

I'm trying to learn about ARM assembler programming using the GNU assembler. I've setup my PC with QEmu and have a Debian ARM-HF chroot environment.
If I assemble and link my test program:
.text
.global _start
_start:
mov r0, #6
bx lr
with:
as test.s -o test.o
ld test.o -o test
Then load the file into gdb and set a breakpoint on _start:
root#Latitude-E6420:/root# gdb test
GNU gdb (GDB) 7.6.1 (Debian 7.6.1-1)
Copyright (C) 2013 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "arm-linux-gnueabihf".
For bug reporting instructions, please see:
...
Reading symbols from /root/test...(no debugging symbols found)...done.
(gdb) break _start
Breakpoint 1 at 0x8054
(gdb)
How do I single step the code, display the assembler source code and monitor the registers?
I tried some basic commands and they did not work:
(gdb) break _start
Breakpoint 1 at 0x8054
(gdb) info regi
The program has no registers now.
(gdb) stepi
The program is not being run.
(gdb) disas
No frame selected.
(gdb) r
Starting program: /root/test
qemu: Unsupported syscall: 26
qemu: uncaught target signal 11 (Segmentation fault) - core dumped
qemu: Unsupported syscall: 26
During startup program terminated with signal SIGSEGV, Segmentation fault.
(gdb)
Your problem here is that you're trying to run an ARM gdb under QEMU's user-mode emulation. QEMU doesn't support the ptrace syscall (that's what syscall number 26 is), so this is never going to work.
What you need to do is run your test binary under QEMU with the QEMU options to enable QEMU's own builtin gdb stub which will listen on a TCP port. Then you can run a gdb compiled to run on your host system but with support for ARM targets, and tell that to connect to the TCP port.
(Emulating ptrace within QEMU is technically very tricky, and it would not provide much extra functionality that you can't already achieve via the QEMU builtin gdbstub. It's very unlikely it'll ever be implemented.)
Minimal working QEMU user mode example
I was missing the -fno-pie -no-pie options:
sudo apt-get install gdb-multiarch gcc-arm-linux-gnueabihf qemu-user
printf '
#include <stdio.h>
#include <stdlib.h>
int main() {
puts("hello world");
return EXIT_SUCCESS;
}
' > hello_world.c
arm-linux-gnueabihf-gcc -fno-pie -ggdb3 -no-pie -o hello_world hello_world.c
qemu-arm -L /usr/arm-linux-gnueabihf -g 1234 ./hello_world
On another terminal:
gdb-multiarch -q --nh \
-ex 'set architecture arm' \
-ex 'set sysroot /usr/arm-linux-gnueabihf' \
-ex 'file hello_world' \
-ex 'target remote localhost:1234' \
-ex 'break main' \
-ex continue \
-ex 'layout split'
;
This leaves us at main, in a split code / disassembly view due to layout split. You will also interested in:
layout regs
which shows the registers.
At the end of the day however, GDB Dashboard is more flexible and reliable: gdb split view with code
-fno-pie -no-pie is required because the packaged Ubuntu GCC uses -fpie -pie by default, and those fail due to a QEMU bug: How to GDB step debug a dynamically linked executable in QEMU user mode?
There was no gdbserver --multi-like functionality for the QEMU GDB stub on QEMU 2.11: How to restart QEMU user mode programs from the GDB stub as in gdbserver --multi?
For those learning ARM assembly, I am starting some runnable examples with assertions and using the C standard library for IO at: https://github.com/cirosantilli/arm-assembly-cheat
Tested on Ubuntu 18.04, gdb-multiarch 8.1, gcc-arm-linux-gnueabihf 7.3.0, qemu-user 2.11.
Freestanding QEMU user mode example
This analogous procedure also works on an ARM freestanding (no standard library) example:
printf '
.data
msg:
.ascii "hello world\\n"
len = . - msg
.text
.global _start
_start:
/* write syscall */
mov r0, #1 /* stdout */
ldr r1, =msg /* buffer */
ldr r2, =len /* len */
mov r7, #4 /* Syscall ID. */
swi #0
/* exit syscall */
mov r0, #0 /* Status. */
mov r7, #1 /* Syscall ID. */
swi #0
' > hello_world.S
arm-linux-gnueabihf-gcc -ggdb3 -nostdlib -o hello_world -static hello_world.S
qemu-arm -g 1234 ./hello_world
On another terminal:
gdb-multiarch -q --nh \
-ex 'set architecture arm' \
-ex 'file hello_world' \
-ex 'target remote localhost:1234' \
-ex 'layout split' \
;
We are now left at the first instruction of the program.
QEMU full system examples
Linux kernel: How to debug the Linux kernel with GDB and QEMU?
Bare metal: https://github.com/cirosantilli/newlib-examples/tree/f70f8a33f8b727422bd6f0b2975c4455d0b33efa#gdb
Single step of an assembly instruction is done with stepi. disas will disassemble around the current PC. info regi will display the current register state. There are some examples for various processors on my blog for my ELLCC cross development tool chain project.
You should add the -g option too to the assembling. Otherwise the codeline info is not included.
That crash probably comes from running some garbage after the code lines.
Maybe you should add the exit system call:
mov eax, 1 ; exit
mov ebx, 0 ; returm value
int 0x80 ; system call

Resources