2

I'm using updated Kali and compiling this for 64- bit

Working through Learning Linux Binary Analysis,'simple ptrace-based debugger' on page 57.

Source code and testfile to debug included.

Problem is accessing the return value of lookup_symbol from the shared object file 'test'.

Regarding the given program: The only thing I've changed is the error check that checks for ET_EXEC, I changed it to ET_DYN so that I can try to trace shared object files, which seem unavoidable when I import stdio.h for the print function. Also, changed the part that checks for ELF files, used use:

h.mem[0] != 0x7f || strcmp((char *)&h.mem1, "ELF" But I don't think the second part is necessary, and my computer spits out an error unless I use "ELF\002\001\001" in the strcmp.

The work of the parent program is done in the lookup_symbol function which searches for the symbol name in the string table and returns the symbol table st_value (the virtual address) of the desired symbol.

From TIS ELF Specification

In executable and shared object files, st_value holds a virtual address. To make these files' symbols more useful for the dynamic linker, the section offset (file interpretation) gives way to a virtual address (memory interpretation) for which the section number is irrelevant.

To search for print_string in 'test':

./tracer ./test print_string

The goal is to break at each print_string call and print the registers at that point, then any key can be pressed to continue execution.

When running the source code the lookup_symbol function returns 0x6b0 on my comp for the value of symtab->st_value and assigns it to h.symaddr.

0x6b0 is the virtual address (value of) of print_string in the symbol table, confirmed by checking readelf

58: 00000000000006b0 27 FUNC GLOBAL DEFAULT 14 print_string

The original from the book uses 0x6b0 directly in the following function:

if ((orig = ptrace(PTRACE_PEEKTEXT, pid, h.symaddr, NULL)) < 0)

This function fails with an error on my computer when it attempts to use h.symaddr=0x6b0:

Beginning analysis of pid: 2462 at 6b0
PTRACE_PEEKER: Input/output error
hello 1
hello 2

The problem is not that it's a virtual address being returned, because as quoted above, both executables and shared object files contain virtual addresses in symbol_table->st_value. I'm not sure why but there is a difference in how this ptracing parent program treats shared objects and executables.

This is the test file:

#include <stdio.h>

void print_string(char * str)
{
    printf("%s\n", str); 
}

int main(int argc, char ** argv)
{
    print_string("hello 1");
    print_string("hello 2");
    return 0;
}

This is the source file:

#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <unistd.h>
#include <fcntl.h>
#include <errno.h>
#include <signal.h>
#include <elf.h>
#include <sys/types.h>
#include <sys/user.h>
#include <sys/stat.h>
#include <sys/ptrace.h>
#include <sys/mman.h>
#include <sys/wait.h>




typedef struct handle
{
    Elf64_Ehdr *ehdr;
    Elf64_Phdr *phdr;
    Elf64_Shdr *shdr;
    uint8_t *mem;
    char *symname;
    Elf64_Addr symaddr;
    struct user_regs_struct pt_reg;
    char *exec;
} handle_t;

Elf64_Addr lookup_symbol(handle_t *, const char *);



int main(int argc, char **argv, char **envp)
{

    int fd;
    handle_t h;
    struct stat st;
    long trap, orig;
    int status, pid;
    char * args[2]; 

    if (argc < 3)
    {
        printf("Usage: %s <program> <function>\n", argv[0]);
        exit(0);
    }


    if ((h.exec = strdup(argv[1])) == NULL)
    {
        perror("strdup"); exit(-1);
    }

    args[0] = h.exec;

    args[1] = NULL;

    if ((h.symname = strdup(argv[2])) == NULL)
    {
        perror("strdup");
        exit(-1);
    }

    if ((fd = open(argv[1], O_RDONLY)) < 0)
    {
        perror("open");
        exit(-1);
    }

    if (fstat(fd, &st) < 0) 
    {
        perror("fstat");
        exit(-1);
    }

    h.mem = mmap(NULL, st.st_size, PROT_READ, MAP_PRIVATE, fd, 0);

    if (h.mem == MAP_FAILED)
    {
        perror("mmap");
        exit(-1);
    }


    h.ehdr = (Elf64_Ehdr *)h.mem;

    h.phdr = (Elf64_Phdr *)(h.mem + h.ehdr->e_phoff);

    h.shdr = (Elf64_Shdr *)(h.mem + h.ehdr->e_shoff);

    if (h.mem[0] != 0x7f)
    {
        printf("%s is not an ELF file\n", h.exec);
        exit(-1);
    }

    if (h.ehdr->e_type != ET_DYN)
    {
        printf("%s is not an ELF dynamic\n", h.exec);
        exit(-1);
    }

    if (h.ehdr->e_shstrndx == 0 || h.ehdr->e_shoff == 0 || h.ehdr->e_shnum == 0)
    {
        printf("Section header table not found\n");
        exit(-1);
    }

    if ((h.symaddr = lookup_symbol(&h, h.symname)) == 0)
    {
        printf("Unable to find symbol: %s not found in executable\n", h.symname);
        exit(-1);
    }

    close(fd);  

    if ((pid = fork()) < 0)
    {
        perror("fork");
        exit(-1);
    }

    if (pid == 0)
    {

        if (ptrace(PTRACE_TRACEME, pid, NULL, NULL) < 0)
        {
            perror("PTRACE_TRACEME");
            exit(-1);
        }

        execve(h.exec, args, envp);
        exit(0);
    }

    wait(&status);
    printf("Beginning analysis of pid: %d at %lx\n", pid, h.symaddr);


    if ((orig = ptrace(PTRACE_PEEKTEXT, pid, h.symaddr, NULL)) < 0)
    {
        perror("PTRACE_PEEKER");
        exit(-1);
    }
    trap = (orig & ~0xff | 0xcc);


    printf("Made it here");


    if (ptrace(PTRACE_POKETEXT, pid, h.symaddr, trap) < 0)
    {
        perror("PTRACE_POKETEXT");
        exit(-1);
    }
trace:

    if (ptrace(PTRACE_CONT, pid, NULL, NULL) < 0)
    {
        perror("PTRACE_CONT");
        exit(-1);
    }
    wait(&status);


    if (WIFSTOPPED(status) && WSTOPSIG(status) == SIGTRAP)
    {
        if (ptrace(PTRACE_GETREGS, pid, NULL, &h.pt_reg) < 0)
        {
            perror("PTRACE_GETREGS");
            exit(-1);
        }

        printf("\nExecutable %s (pid: %d) has hit breakpoint 0x%lx\n", h.exec, pid, h.symaddr);

    printf("%%rcx: %llx\n%%rdx: %llx\n%%rbx: %llx\n" 
                "%%rax: %llx\n%%rdi: %llx\n%%rsi: %llx\n" 
                "%%r8: %llx\n%%r9: %llx\n%%r10: %llx\n%%" 
                "%%r11: %llx\n%%r12: %llx\n%%r13: %llx\n%%" 
                "%%r14: %llx\n%%r15: %llx\n%%rsp: %llx\n%%");
    printf("\nHit any key to continue: ");
    getchar();
    if (ptrace(PTRACE_POKETEXT, pid, h.symaddr, orig) < 0)
    {
        perror("PTRACE_POKETEXT");
        exit(-1);
    }

    h.pt_reg.rip = h.pt_reg.rip - 1;
    if (ptrace(PTRACE_SETREGS, pid, NULL, &h.pt_reg) < 0)
    {
        perror("PTRACE_SETREGS");
        exit(-1);
    }

    if (ptrace(PTRACE_SINGLESTEP, pid, NULL, NULL) < 0)
    {
        perror("PTRACE_SINGLESTEP");
        exit(-1);
    }
    wait(NULL);

    if (ptrace(PTRACE_POKETEXT, pid, h.symaddr, trap) < 0)
    {
        perror("PTRACE_POKETEXT");
        exit(-1);
    }

    goto trace;
    }

    if (WIFEXITED(status)) 
        printf("Completed tracing pid: %d\n", pid);

    exit(0);
}




    Elf64_Addr lookup_symbol(handle_t *h, const char *symname)
    {
        int i, j;
        char *strtab;
        Elf64_Sym *symtab;
        for (i = 0; i < h->ehdr->e_shnum; i++)      
        {
            if (h->shdr[i].sh_type == SHT_SYMTAB)   
            {
                strtab = (char *)&h->mem[h->shdr[h->shdr[i].sh_link].sh_offset];

                symtab = (Elf64_Sym *)&h->mem[h->shdr[i].sh_offset];


                for (j = 0; j < h->shdr[i].sh_size/sizeof(Elf64_Sym); j++)
                {
                    if (strcmp(&strtab[symtab->st_name], symname) == 0)
                        //printf("symtab->st_value is 0x%lx\n", symtab->st_value);  
                        return (symtab->st_value);
                    symtab++;
                }
            }
        }
        return 0;
    }
flerb
  • 131
  • 7

2 Answers2

1

The problem is that when I compile test.c with gcc, gcc creates a shared object file.

This is the case with even the simplest of programs: //simplereturn.c int main(void) { return 22; }

from readelf:

Type:                              DYN (Shared object file)

When I try to parse this with tracer, previously I had changed the error-checking section that ensured that the file that I'm parsing with ptrace is ET_EXEC (an executable file). I changed it so it checked if the traced file was ET_DYN (shared object file).

Doing this allowed the error-check to pass for the simple C programs that I wanted to trace, but had unexpected (at least for me) consequences. When lookup_symbol was returning it was in fact returning a virtual address (0x6b0), which agrees with the symbol table of the readelf, and the function address in objdump -D. Can even check it in gdb. However, interestingly, a break at print_string is placed at 0x6bc, while the start of the print_string function is at 0x6b0. Probably not irrelevant.

After a lot of research and struggle, the problem was an easy fix.

I created an executable instead of a shared object file. The requisite as far as I can tell is that it uses no external libraries or functions, or I guess probably that it's self-contained. Either way, gcc won't make them.

;test.asm
section .text
    global _start

_start:
    mov rsi, hello1
    call print_string
    mov rsi, hello2
    call print_string

    xor rdi,rdi
    mov rax, 60
    syscall

print_string:
    mov rax, 1
    mov rdi, 1
    mov rdx, 6
    syscall
    ret

section .data
    hello1 db "Hello1"
    hello2 db "Hello2"

Compiled with nasm -g -f elf64 test.asm && ld -m elf_x86_64 -o testasm testasm.o

Now readelf -a testasm confirms that it's an executable

ELF Header:
  Magic:   7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 
  Class:                             ELF64
  Data:                              2's complement, little endian
  Version:                           1 (current)
  OS/ABI:                            UNIX - System V
  ABI Version:                       0
 Type:                              EXEC (Executable file) //EXEC NOW
  Machine:                           Advanced Micro Devices X86-64
  Version:                           0x1
  Entry point address:               0x4000b0
  Start of program headers:          64 (bytes into file)
  Start of section headers:          616 (bytes into file)
  Flags:                             0x0
  Size of this header:               64 (bytes)
  Size of program headers:           56 (bytes)
  Number of program headers:         2
  Size of section headers:           64 (bytes)
  Number of section headers:         6
  Section header string table index: 5

And I switched the error checking in the original back:

#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <unistd.h>
#include <fcntl.h>
#include <errno.h>
#include <signal.h>
#include <elf.h>
#include <sys/types.h>
#include <sys/user.h>
#include <sys/stat.h>
#include <sys/ptrace.h>
#include <sys/mman.h>
#include <sys/wait.h>


typedef struct handle
{
    Elf64_Ehdr *ehdr;
    Elf64_Phdr *phdr;
    Elf64_Shdr *shdr;
    uint8_t *mem;
    char *symname;
    Elf64_Addr symaddr;
    struct user_regs_struct pt_reg;
    char *exec;
} handle_t;

Elf64_Addr lookup_symbol(handle_t *, const char *);

int main(int argc, char **argv, char **envp)
{
    int fd;
    handle_t h;
    struct stat st;
    long trap, orig;
    int status, pid;
    char * args[2];

    if (argc < 3)
    {
        printf("Usage: %s <program> <function>\n", argv[0]);
        exit(0);
    }

    if ((h.exec = strdup(argv[1])) == NULL)
    {
        perror("strdup"); exit(-1);
    }

    args[0] = h.exec;

    args[1] = NULL;

    if ((h.symname = strdup(argv[2])) == NULL)
    {
        perror("strdup");
        exit(-1);
    }

    if ((fd = open(argv[1], O_RDONLY)) < 0)
    {
        perror("open");
        exit(-1);
    }

    if (fstat(fd, &st) < 0) 
    {
        perror("fstat");
        exit(-1);
    }


    h.mem = mmap(NULL, st.st_size, PROT_READ, MAP_PRIVATE, fd, 0);


    if (h.mem == MAP_FAILED)
    {
        perror("mmap");
        exit(-1);
    }


    h.ehdr = (Elf64_Ehdr *)h.mem;

    h.phdr = (Elf64_Phdr *)(h.mem + h.ehdr->e_phoff);

    h.shdr = (Elf64_Shdr *)(h.mem + h.ehdr->e_shoff);

    if ((h.mem[0] != 0x7f) || strcmp((char *)&h.mem[1], "ELF\002\001\001"))
    {
        printf("%s is not an ELF file\n", h.exec);
        exit(-1);
    }

    if (h.ehdr->e_type != ET_EXEC) 
    {
        printf("%s is not an ELF executable\n", h.exec);
        exit(-1);
    }


    if (h.ehdr->e_shstrndx == 0 || h.ehdr->e_shoff == 0 || h.ehdr->e_shnum == 0)
    {
        printf("Section header table not found\n");
        exit(-1);
    }

    if ((h.symaddr = lookup_symbol(&h, h.symname)) == 0)
    {
        printf("Unable to find symbol: %s not found in executable\n", h.symname);
        exit(-1);
    }


    close(fd);  


    if ((pid = fork()) < 0)
    {
        perror("fork");
        exit(-1);
    }

    if (pid == 0)
    {
        if (ptrace(PTRACE_TRACEME, pid, NULL, NULL) < 0)
        {
            perror("PTRACE_TRACEME");
            exit(-1);
        }
        execve(h.exec, args, envp);
        exit(0);
    }




    wait(&status);
    printf("Beginning analysis of pid: %d at %lx\n", pid, h.symaddr);

    if ((orig = ptrace(PTRACE_PEEKTEXT, pid, h.symaddr, (void *)0)) < 0)
    {
        perror("PTRACE_PEEKER");
        exit(-1);
    }
    trap = (orig & ~0xff | 0xcc);

    if (ptrace(PTRACE_POKETEXT, pid, h.symaddr, trap) < 0)
    {
        perror("PTRACE_POKETEXT");
        exit(-1);
    }
trace:
    if (ptrace(PTRACE_CONT, pid, NULL, NULL) < 0)
    {
        perror("PTRACE_CONT");
        exit(-1);
    }
    wait(&status);

    if (WIFSTOPPED(status) && WSTOPSIG(status) == SIGTRAP)
    {
        if (ptrace(PTRACE_GETREGS, pid, NULL, &h.pt_reg) < 0)
        {
            perror("PTRACE_GETREGS");
            exit(-1);
        }

        printf("\nExecutable %s (pid: %d) has hit breakpoint 0x%lx\n", h.exec, pid, h.symaddr);

    printf("%%rcx: %llx\n%%rdx: %llx\n%%rbx: %llx\n" 
                "%%rax: %llx\n%%rdi: %llx\n%%rsi: %llx\n" 
                "%%r8: %llx\n%%r9: %llx\n%%r10: %llx\n%%" 
                "%%r11: %llx\n%%r12: %llx\n%%r13: %llx\n%%" 
                "%%r14: %llx\n%%r15: %llx\n%%rsp: %llx\n%%");
    printf("\nHit any key to continue: ");
    getchar();
    if (ptrace(PTRACE_POKETEXT, pid, h.symaddr, orig) < 0)
    {
        perror("PTRACE_POKETEXT");
        exit(-1);
    }

    h.pt_reg.rip = h.pt_reg.rip - 1;

    if (ptrace(PTRACE_SETREGS, pid, NULL, &h.pt_reg) < 0)
    {
        perror("PTRACE_SETREGS");
        exit(-1);
    }

    if (ptrace(PTRACE_SINGLESTEP, pid, NULL, NULL) < 0)
    {
        perror("PTRACE_SINGLESTEP");
        exit(-1);
    }
    wait(NULL);

    if (ptrace(PTRACE_POKETEXT, pid, h.symaddr, trap) < 0)
    {
        perror("PTRACE_POKETEXT");
        exit(-1);
    }

    goto trace;
    }

    if (WIFEXITED(status)) 
        printf("Completed tracing pid: %d\n", pid);

    exit(0);
}




    Elf64_Addr lookup_symbol(handle_t *h, const char *symname)
    {
        int i, j;
        char *strtab;
        Elf64_Sym *symtab;
        for (i = 0; i < h->ehdr->e_shnum; i++)  
        {
            if (h->shdr[i].sh_type == SHT_SYMTAB)
            {
                strtab = (char *)&h->mem[h->shdr[h->shdr[i].sh_link].sh_offset];

                symtab = (Elf64_Sym *)&h->mem[h->shdr[i].sh_offset];

                for (j = 0; j < h->shdr[i].sh_size/sizeof(Elf64_Sym); j++)
                {
                    if (strcmp(&strtab[symtab->st_name], symname) == 0)
                        return (symtab->st_value);
                    symtab++;
                }
            }
        }
        return 0;
    }

And now when run it it works $ ./tracer ./testasm print_string

Beginning analysis of pid: 5378 at 4000d8

Executable ./testasm (pid: 5378) has hit breakpoint 0x4000d8
%rcx: 564ed7b07050
%rdx: 7fb9525a9760
%rbx: 7fffffc2
%rax: 1
%rdi: 3e
%rsi: 1958ac0
%r8: 7ffcd2b98848
%r9: 7ffcd2b98828
%r10: 3d2bea1a8
%%r11: 564ed7b07010
%r12: 0
%r13: 7fb9527d2728
%%r14: 57f00000001
%r15: 801
%rsp: 203bec
%
Hit any key to continue: 
Hello1
Executable ./testasm (pid: 5378) has hit breakpoint 0x4000d8
%rcx: 564ed7b07050
%rdx: 7fb9525a9760
%rbx: 7fffffc2
%rax: 1
%rdi: 3e
%rsi: 1958ac0
%r8: 7ffcd2b98848
%r9: 7ffcd2b98828
%r10: 3d2bea1a8
%%r11: 564ed7b07010
%r12: 0
%r13: 7fb9527d2728
%%r14: 57f00000001
%r15: 801
%rsp: 203bec
%
Hit any key to continue: 
Hello2Completed tracing pid: 5378

checking gdb and readelf Lookup symbol returns value at 0x4000d8, as the symbol table from readelf -a testasm says it should:

Symbol table '.symtab' contains 11 entries:
   Num:    Value          Size Type    Bind   Vis      Ndx Name
     0: 0000000000000000     0 NOTYPE  LOCAL  DEFAULT  UND 
     1: 00000000004000b0     0 SECTION LOCAL  DEFAULT    1 
     2: 00000000006000ec     0 SECTION LOCAL  DEFAULT    2 
     3: 0000000000000000     0 FILE    LOCAL  DEFAULT  ABS test.asm
     4: 00000000004000d8     0 NOTYPE  LOCAL  DEFAULT    1 print_string
     5: 00000000006000ec     0 NOTYPE  LOCAL  DEFAULT    2 hello1
     6: 00000000006000f2     0 NOTYPE  LOCAL  DEFAULT    2 hello2
     7: 00000000004000b0     0 NOTYPE  GLOBAL DEFAULT    1 _start
     8: 00000000006000f8     0 NOTYPE  GLOBAL DEFAULT    2 __bss_start
     9: 00000000006000f8     0 NOTYPE  GLOBAL DEFAULT    2 _edata
    10: 00000000006000f8     0 NOTYPE  GLOBAL DEFAULT    2 _end

So the differences between tracing an executable or shared object file in this case were what broke it.

flerb
  • 131
  • 7
0

You can build without dynamic linking like this:

gcc -static -o test test.c

Then it'll create DT_EXEC type executable not DT_DYN.

wonbear
  • 1
  • 2