Chapter 3 - System Calls

System Calls

• Will always switch to Kernel mode
• Syscall number is fixed but user identify it by syscall func name.

How sys call works

• When userspace programs call syscall, then they store a number corresponsing to syscall on a register %eax
• The wrapper that has this mov command to register will also include a trap called 0x80 . In most newer architecture sysenter is used to enter kernel mode.
• When this trap is encountered, CPU will start the system_call() subroutine located in arch/i386/entry.S see.
  • this saves the register value to kernel stack
  • checks the validity of syscall and executes it
  • returns the value to program stack and switches to user mode
• sys_call_table has all the list of system calls and its numbers.
• for example - sys_execve syscall has number 11 and hence in sys_call_table 11th entry store the address of routine for sys_execve handler.
• use strace to trace syscalls in a program.

library functions

• wrappers around syscall.
• Most common Linux is implemenetd in GNU C
• glibc version is determined from /lib/libc.so.6
• one can also perform list dynamic list dependency like ldd myprog | grep libc to check the version of libc

#include <gnu/libc-version.h>
const char *gnu_get_libc_version(void);

One can also call gnu_get_libc_version to get this.

#include <stdio.h>
#include <gnu/libc-version.h>
#include <errno.h>

int main() {
    const char* version_no = gnu_get_libc_version();
    if !(version_no) {
      fprintf(stderr, "Error: Failed to retrieve glibc version\n");
      return 1;
    }
    printf("version number :- %s\n", version_no);
    return 0;
}

and output

Anurag252 ➜ /workspaces/codespaces-blank $ gcc main.c ; ./a.out
version number :- 2.39

error handling from syscall and library functions

• a few prgms never fail like getpid and exit and their return need not be checked
  • system call failures set errno
  • perror prints an error message pointed by msg arguement and then errno error
  • strerror returns error corresponsing to errono.

#include <stdio.h>
#include <errno.h>
#include <unistd.h>
#include <string.h>
#include <fcntl.h>

int main() {
   ssize_t read_bytes = open("some/file", 1);
   if (read_bytes == -1) {
        perror("some error occured:- ");
        printf("%s\n", strerror(errno));
        return 1;
   }
   return 0;

}

• std lib function
  • some return errno, some do not set errno but return an int, other do not do both

    atoi(), atol(), and strtol()
    

feature flags

• one can define feature flags like this

#include <stdio.h>
int main() {

  if (MY_FEATURE) {
      printf("hello\n");
      return 0;
  } else {
      printf("bye\n");
      return 1;
  }

}

to test

gcc -D MY_FEATURE=0 main.c ; ./a.out

• there are pre-defined FFs which enable and disable multiple certain functionality and specially platform related code
  • _POSIX_SOURCE
  • _POSIX_C_SOURCE
  • _XOPEN_SOURCE

system data type

• there are some pre-defined type sszie_t from open which is int but to capture platform related specifics, they are system defined
• generally have _t at the end and it appears as typedef int pid_t;
• most available in <sys/types.h>

Exercise

man 2 reboot
This  system  call  fails  (with  the  error  EINVAL)  unless magic equals LINUX_REBOOT_MAGIC1 (that is, 0xfee1dead) and magic2 equals LINUX_RE‐
       BOOT_MAGIC2 (that is, 0x28121969).  However, since Linux 2.1.17 also LINUX_REBOOT_MAGIC2A (that is, 0x05121996)  and  since  Linux  2.1.97  also
       LINUX_REBOOT_MAGIC2B  (that  is,  0x16041998) and since Linux 2.5.71 also LINUX_REBOOT_MAGIC2C (that is, 0x20112000) are permitted as values for
       magic2.  (The hexadecimal values of these constants are meaningful.)

These hexadecimals are birrthdates 0x28121969 - 28.12.1969 and so on