Calling Assembly from C (64 bit)

27 Nov 2012

In a previous post, I spoke about calling assembly language routines from C in 32-bit land. In this post, I aim to show you the same technique only this time using a 64 bit architecture.

I’ll be using the same technology stack as last time with NASM as my assembler and GCC as my C compiler. The code snippets that you’ll see in this post have been compiled and linked together on the linux platform.

Alright then. On with the code!

The C-code is not different from the 32-bit version as mentioned in my previous post. Now that’s portability for you!

#include <stdio.h>

/** Forward declaration for our add function */
int add(int, int);

int main() {

   /* add some numbers */
   int x = add(5, 4);

   /* print the result out */
   printf("5+4=%d\n", x);

   return 0;
}

The assembly language module has changed somewhat though.

global add

section .text

add:
   mov   rax, rdi    ; get the first parameter
   add   rax, rsi    ; add the second parameter
                     ; leaving the return value in rax
   ret

Immediately you can see that a new set of registers are being used and the stack is no longer being used to take parameters in from external calls.

The change in calling convention can be studied in further detail here, however it’s just important to mention this:

Once arguments are classified, the registers get assigned (in left-to-right order) for passing as follows: If the class is MEMORY, pass the argument on the stack. If the class is INTEGER, the next available register of the sequence %rdi, %rsi, %rdx, %rcx, %r8 and %r9 is used

So, the first argument is passed in RDI and the second in RSI. The return value remains the same being passed back to the caller using the accumulator register (RAX in 64-bit land). There are many resources to read up on when it comes to the 64 bit architecture; Intel and Wikipedia are both excellent resources.

Finally, when it comes to building these files together, it’s (again) very similar to its 32bit counterpart.

$ gcc -c add.c -o add.o
$ nasm -f elf64 maths.asm -o maths.o
$ gcc -m64 add.o maths.o -o add

We’re done!

Unix Networking: Server

26 Nov 2012

The following snippet is a bare-bones server.

int sockfd, new_fd;
struct addrinfo hints, *servinfo, *p;
struct sockaddr_storage their_addr;
socklen_t sin_size;
struct sigaction sa;
int yes=1;
char s[INET6_ADDRSTRLEN];
int rv;

/* fill out the address structure */
memset(&hints, 0, sizeof hints);
hints.ai_family = AF_UNSPEC;
hints.ai_socktype = SOCK_STREAM;
hints.ai_flags = AI_PASSIVE;

/* get the address information */
getaddrinfo(NULL, PORT, &hints, &servinfo);

/* open the socket */
sockfd = socket(p->ai_family, p->ai_socktype, p->ai_protocol);

/* allow the socket to be re-used */
setsockopt(sockfd, SOL_SOCKET, SO_REUSEADDR, &yes, sizeof(int));

/* bind the socket to the address */
bind(sockfd, p->ai_addr, p->ai_addrlen);

/* free up the address structure */
freeaddrinfo(servinfo);

/* start the socket listening */
listen(sockfd, BACKLOG);

/* accept the first connect */
new_fd = accept(sockfd, (struct sockaddr *)&their_addr, &sin_size);

/* send the client some data */
send(new_fd, "Hello, world!", 13, 0);

/* finished with the client & server */
close(new_fd);
close(sockfd);

Further reading The getaddrinfo system call The socket system call The bind system call The listen system call The accept system call The recv system call The send system call

Unix Networking: Client

26 Nov 2012

The following snippet outlines the skeleton of a client socket application. It’s the bare bones of what is needed to establish the client connection.

int sockfd, numbytes;
char buf[MAXDATASIZE];
struct addrinfo hints, *servinfo, *p;
int rv;
char s[INET6_ADDRSTRLEN];

/* fill out the address info */
memset(&hints, 0, sizeof hints);
hints.ai_family = AF_UNSPEC;
hints.ai_socktype = SOCK_STREAM;

/* create the actual server address */
getaddrinfo("www.remoteplace.com", PORT, &hints, &servinfo);

/* create the socket file descriptor */
sockfd = socket(p->ai_family, p->ai_socktype, p->ai_protocol);

/* perform the network connect */
connect(sockfd, p->ai_addr, p->ai_addrlen);

/* free up the addres structure */
freeaddrinfo(servinfo);

/* receive some data on the socket */
recv(sockfd, buf, MAXDATASIZE-1, 0));

/* close the socket off */
close(sockfd);

###Further reading The getaddrinfo system call The socket system call The connect system call The recv system call The send system call

Using the GNU Debugger

25 Nov 2012

Bundled with the GNU toolset is a full featured debugger in GDB. This blog post aims to be a cheatsheet to get yourself around this tool quickly.

The example program

int main(int argc, char *argv[]) {
  /* some local variables */
  int x = 1, y = 2, z = 0;

  /* something that will go bad */
  int ans = x / z;

  return 0;
}

Compile your code with debug info

You need to use the -g switch with GCC in order to compile debug symbols. This helps the feedback that GDB will give you a great deal.

$ gcc -g test.c -o test

Loading your program into the debugger

So, in our example above the name of our executable would be “test”. So, we just tell gdb to load test by passing it in.

$ gdb test

You can even start gdb with a text user interface, so I’ve just discovered.

$ gdb test -tui

It looks like this - not bad, eh?

GDB Screenshot

Now we instruct gdb that we’d like to run the application

(gdb) run
Starting program: /home/michael/Development/gdb-tute/test 
Program received signal SIGFPE, Arithmetic exception.
0x00000000004004c3 in main (argc=1, argv=0x7fffffffe4d8) at test.c:9
9    ans = x / z;

So you can see here that dividing by zero wasn’t the best career move for this program. But It nicely let us know that there was an arithmetic exception on line 9!

Examining data

To get just the call stack information, we issue the “backtrace” instruction

(gdb) backtrace
#0  0x00000000004004c3 in main (argc=1, argv=0x7fffffffe4d8) at test.c:9

Examining data with print. We can view the value of variables by passing them to the print statement.

(gdb) print x
$1 = 1
(gdb) print y
$2 = 2
(gdb) print z
$3 = 0

Print also supports printf style specifiers. Examining data at an address. We can view the value of data at a memory location by using the x command.

(gdb) print &z
$4 = (int *) 0x7fffffffe3e8
(gdb) x 0x7fffffffe3e8
0x7fffffffe3e8: 0x00000000

Setting variables while attached is done with set. We can stop the error in the program at runtime using this mixed with a breakpoint.

Breakpoint 1, main (argc=1, argv=0x7fffffffe4d8) at test.c:6
6    int x = 1, y = 2, z = 0;
(gdb) next
7    int ans = 0;
(gdb) set (z = 5)
(gdb) print z
$1 = 5

Working with breakpoints

To set a breakpoint in gdb, just pass the name of what you’d like to break on to the break command.

(gdb) break main
Breakpoint 1 at 0x40049f: file test.c, line 6.
(gdb) run
Starting program: /home/michael/Development/gdb-tute/test 
Breakpoint 1, main (argc=1, argv=0x7fffffffe4d8) at test.c:6
6    int x = 1, y = 2, z = 0;

You can set a breakpoint at a specific source code line number:

(gdb) break 7
Breakpoint 1 at 0x4004b4: file test.c, line 7.

Finally, you can make a breakpoint using a combination of source code file name and line location:

(gdb) break test.c:7
Breakpoint 3 at 0x4004b4: file test.c, line 7.

View your existing breakpoints with info breakpoints

(gdb) info breakpoints
Num     Type           Disp Enb Address            What
3       breakpoint     keep y   0x00000000004004b4 in main at test.c:7

You can clear any set breakpoint with clear.

(gdb) clear main
Deleted breakpoint 1

You can step into code by issuing the step command

(gdb) step
7    int ans = 0;
(gdb) 
9    ans = x / z;
(gdb) 
Program received signal SIGFPE, Arithmetic exception.
0x00000000004004c3 in main (argc=1, argv=0x7fffffffe4d8) at test.c:9
9    ans = x / z;

You step over code by issuing next. Its usage is the same as step.

That’s all for now. As I come across other useful bits from GDB, I’ll certainly post them here. The items above have been lifesavers for me from time to time.

Unix IPC: Working with Signals