arsh/main.c

#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <stdint.h>
#include <stdio.h>
#include <sys/wait.h>
#include <signal.h>
#include <setjmp.h>
#include <ctype.h>

constexpr size_t max_length = 1024;
constexpr size_t max_env_var_lenght = 256;
const char* token_separators = " \t\n\r";

static sigjmp_buf restart_jmp_buf;
static constexpr int restart_value = 69;
static volatile sig_atomic_t jump_active = 0;

void sigint_handler(int signo)
{
  if(!jump_active)
  {
    return;
  }
  siglongjmp(restart_jmp_buf, restart_value);
}

/*Checked calloc: if allocation fails -> exists*/
void* ccalloc(size_t size, size_t n)
{
  void* ret = calloc(size, n);
  if(!ret)
  {
    perror("calloc failed");
    exit(1);
  }
  return ret;
}

static constexpr size_t max_line = 1024;
static char string_buffer[max_line];
char* readline(char* print, FILE* stream)
{
  if(print)
  {
    printf("%s", print);
  }
  if(fgets(string_buffer, max_line, stream) == NULL)
  {
    if (feof(stdin))
    {
        return NULL;   // Ctrl+D
    }
    if(ferror(stream))
    {
      perror("fgets error");
      clearerr(stdin);
      return NULL;
    }
    return NULL;
  }
  
  size_t length = strlen(string_buffer);
  //Remove trailing \n
  if (length > 0 && string_buffer[length - 1] == '\n')
  {
    string_buffer[length - 1] = '\0';
  }
  return string_buffer;
}

/**
 * Struct to represent a command and its arguments.
 */
struct command
{
  /**
   * IO redirections; redirect[i] should be used as fd i in the child.
   * A value of -1 indicates no redirect.
   */
  int redirect[2];
  /** The arguments; must be NULL-terminated. */
  char* argv[];
};
/** The name of the executable. */
char* command_name(struct command* cmd)
{
  return cmd->argv[0];
}

/**
 * Struct to represent a pipeline of commands. The intention is that cmd[i]'s
 * output goes to cmd[i+1]'s input.
 */
struct pipeline
{
  /** The total number of commands. */
  size_t n_cmds;
  struct command* cmds[];
};

void* parser_memory[max_length] = {0};
size_t parser_memory_index = 0;
void parser_allocated(void* ptr)
{
  parser_memory[parser_memory_index] = ptr;
  if(parser_memory_index == max_length)
  {
    fprintf(stderr, "Not enough memory for parser: Memory will leak!");
    exit(1);
  }
  parser_memory_index++;
}

void* parser_realloc(void* ptr, size_t size)
{
  size_t index;
  for(index = 0; index < parser_memory_index; index++)
  {
    if(parser_memory[index] == ptr)
    {
      break;
    }
  }
  ptr = realloc(ptr, size);
  if (!ptr)
  {
    perror("realloc failed");
    exit(1);
  }
  parser_memory[index] = ptr;
  return ptr;
}

void parser_free()
{
  for(size_t i = 0; i < parser_memory_index; i++)
  {
    free(parser_memory[i]);
  }
  memset(parser_memory, 0, max_length);
  parser_memory_index = 0;
}

char* expand_variable(const char* name)
{
    char* val = getenv(name);
    char* ret = val ? strdup(val) : strdup("");
    parser_allocated(ret);
    return ret;
}

static void append_char(char** buf, size_t* len, size_t* cap, char c)
{
  if (*len + 1 >= *cap)
  {
    *cap *= 2;
    *buf = parser_realloc(*buf, *cap);
  }
  (*buf)[(*len)++] = c;
}

static void append_str(char** buf, size_t* len, size_t* cap, const char* s)
{
  while (*s)
    append_char(buf, len, cap, *s++);
}

struct pipeline* parse_pipeline(char *str);
int execute_pipeline(struct pipeline* pl, int capture, char** out);

static char* run_subcommand(const char* cmd)
{
    struct pipeline* pl = parse_pipeline((char*)cmd);

    char* output = NULL;
    execute_pipeline(pl, 1, &output);

    return output;
}

char* parse_token(char** input)
{
  char* p = *input;

  while (*p && isspace(*p))
    p++;

  if (!*p)
  {
    *input = p;
    return NULL;
  }

  size_t cap = 64;
  size_t len = 0;
  char* buf = malloc(cap);
  if (!buf)
  {
    perror("malloc");
    exit(1);
  }

  int in_single = 0;
  int in_double = 0;

  while (*p)
  {
    if (!in_single && !in_double && isspace(*p))
      break;

    if (!in_double && *p == '\'')
    {
      in_single = !in_single;
      p++;
      continue;
    }

    if (!in_single && *p == '"')
    {
      in_double = !in_double;
      p++;
      continue;
    }

    if (!in_single && *p == '\\')
    {
      p++;
      if (*p)
        append_char(&buf, &len, &cap, *p++);
      continue;
    }

    if (!in_single && *p == '$')
    {
      p++;

      if (*p == '(')
      {
	p++;
	size_t cap2 = 256;
	size_t len2 = 0;
	char* cmd = malloc(cap2);
	if (!cmd) { perror("malloc"); exit(1); }

	int depth = 1;
	int in_single2 = 0;
	int in_double2 = 0;

	while (*p && depth > 0)
	{
	  if (!in_double2 && *p == '\'')
	  {
	    in_single2 = !in_single2;
	  }
	  else if (!in_single2 && *p == '"')
	  {
	    in_double2 = !in_double2;
	  }
	  else if (!in_single2 && !in_double2)
	  {
	    if (*p == '(') depth++;
	    else if (*p == ')') depth--;
	  }

	  if (depth > 0)
	    append_char(&cmd, &len2, &cap2, *p);

	  p++;
	}

	if (depth != 0)
	{
	  fprintf(stderr, "syntax error: unclosed $( )\n");
	  free(cmd);
	  return NULL;
	}

	cmd[len2] = '\0';
	
	char* out = run_subcommand(cmd);
	append_str(&buf, &len, &cap, out);

	free(cmd);
	continue;
      }

      if (*p == '{')
      {
        p++;
        char name[max_env_var_lenght];
        int i = 0;
        while (*p && *p != '}')
          name[i++] = *p++;
        name[i] = '\0';
        if (*p == '}') p++;
        char* val = expand_variable(name);
        append_str(&buf, &len, &cap, val);
        continue;
      }

      char name[max_env_var_lenght];
      int i = 0;
      while (*p && (isalnum(*p) || *p == '_'))
        name[i++] = *p++;
      name[i] = '\0';

      char* val = expand_variable(name);
      append_str(&buf, &len, &cap, val);
      continue;
    }

    if (!in_single && *p == '`')
    {
      p++;
      char cmd[max_length];
      int i = 0;
      while (*p && *p != '`')
        cmd[i++] = *p++;
      if (*p == '`') p++;
      cmd[i] = '\0';
      char* out = run_subcommand(cmd);
      append_str(&buf, &len, &cap, out);
      continue;
    }

    append_char(&buf, &len, &cap, *p++);
  }

  if (in_single || in_double)
  {
    fprintf(stderr, "syntax error: unclosed quote\n");
    free(buf);
    *input = p;
    return NULL;
  }

  buf[len] = '\0';
  parser_allocated(buf);

  *input = p;
  return buf;
}

/**
 * Parses str into a freshly allocated command struct and returns a pointer to it.
 * The redirects in the returned command will be set to -1, ie no redirect.
 */
struct command* parse_command(char* str) 
{
  /* Copy the input line in case the caller wants it later. */
  char* copy = strndup(str, max_length);
  parser_allocated(copy);
  char* token;
  int i = 0;

  /*
   * Being lazy and allocating way too much memory for the args array.
   * Using calloc to ensure it's zero-initialised, which is important because
   * execvp expects a NULL-terminated array of arguments.
   */
  struct command* ret = ccalloc(sizeof(struct command) + strlen(copy) * sizeof(char*), 1);
  parser_allocated(ret);

  char* p = copy;

  while (*p)
  {
    char* token = parse_token(&p);

    if (!token)
      break;

    ret->argv[i++] = token;
  }
  ret->redirect[0] = ret->redirect[1] = -1;
  return ret;
}

/**
 * Parses str into a freshly allocated pipeline_struct and returns a pointer to
 * it.  All commands in cmds will also be freshy allocated, and have their
 * redirects set to -1, ie no redirect.
 */
struct pipeline* parse_pipeline(char *str)
{
  char* copy = strndup(str, max_length);
  parser_allocated(copy);

  size_t cap = 4;
  size_t count = 0;
  struct command** cmds = malloc(sizeof(struct command*) * cap);
  if (!cmds)
  {
    perror("malloc");
    exit(1);
  }

  int in_single = 0;
  int in_double = 0;
  int in_backtick = 0;
  int paren_depth = 0;

  char* start = copy;
  char* p = copy;

  while (*p)
  {
    if (!in_double && !in_backtick && *p == '\'')
      in_single = !in_single;

    else if (!in_single && !in_backtick && *p == '"')
      in_double = !in_double;

    else if (!in_single && !in_double && *p == '`')
      in_backtick = !in_backtick;

    else if (!in_single && !in_double && !in_backtick)
    {
      if (*p == '$' && *(p+1) == '(')
      {
        paren_depth++;
        p++; // skip '(' next iteration
      }
      else if (*p == '(' && paren_depth > 0)
      {
        paren_depth++;
      }
      else if (*p == ')' && paren_depth > 0)
      {
        paren_depth--;
      }
      else if (*p == '|' && paren_depth == 0)
      {
        *p = '\0';

        if (count >= cap)
        {
          cap *= 2;
          cmds = realloc(cmds, sizeof(struct command*) * cap);
          if (!cmds)
          {
            perror("realloc");
            exit(1);
          }
        }

        cmds[count++] = parse_command(start);
        start = p + 1;
      }
    }

    p++;
  }

  cmds[count++] = parse_command(start);

  struct pipeline* ret =
    ccalloc(sizeof(struct pipeline) + count * sizeof(struct command*), 1);

  parser_allocated(ret);
  parser_allocated(cmds);
  ret->n_cmds = count;

  for (size_t i = 0; i < count; i++)
    ret->cmds[i] = cmds[i];

  return ret;
}

void close_ALL_the_pipes(int n_pipes, int (*pipes)[2])
{
  for (int i = 0; i < n_pipes; ++i)
  {
    close(pipes[i][0]);
    close(pipes[i][1]);
  }
}

int exec_with_redir(struct command* command, int n_pipes, int (*pipes)[2])
{
  int fd = -1;
  if ((fd = command->redirect[0]) != -1)
  {
    dup2(fd, STDIN_FILENO);
  }
  if ((fd = command->redirect[1]) != -1)
  {
    dup2(fd, STDOUT_FILENO);
  }
  close_ALL_the_pipes(n_pipes, pipes);
  return execvp(command_name(command), command->argv);
}

static struct sigaction s_old;

pid_t run_with_redir(struct command* command, int n_pipes, int (*pipes)[2])
{
  
  pid_t child_pid = fork();
  
  if(child_pid < 0)
  {
    perror("Fork failed");
    exit(1);
  }
  if (child_pid) 
  {  /* We are the parent. */
      return child_pid;
  } 
  else 
  {  // We are the child. */
    sigaction(SIGINT, &s_old, NULL);
    exec_with_redir(command, n_pipes, pipes);
    perror(command_name(command));
    exit(1);
  }
}

int cd(char* path)
{
  return chdir(path);
}

int execute_pipeline(struct pipeline* pl, int capture, char** out)
{
    int n_pipes = pl->n_cmds - 1;
    int (*pipes)[2] = ccalloc(sizeof(int[2]), n_pipes);
    parser_allocated(pipes);

    for (int i = 1; i < pl->n_cmds; ++i)
    {
        pipe(pipes[i-1]);
        pl->cmds[i]->redirect[STDIN_FILENO] = pipes[i-1][0];
        pl->cmds[i-1]->redirect[STDOUT_FILENO] = pipes[i-1][1];
    }

    int capture_pipe[2];

    if (capture)
    {
        pipe(capture_pipe);
        pl->cmds[pl->n_cmds - 1]->redirect[STDOUT_FILENO] = capture_pipe[1];
    }

    pid_t* pids = ccalloc(sizeof(pid_t), pl->n_cmds);
    parser_allocated(pids);

    for (int i = 0; i < pl->n_cmds; ++i)
    {
      struct command* cmd = pl->cmds[i];
      char* cmd_name = command_name(cmd);
      if(!cmd_name)
      {
	continue;
      }
      if(strlen(cmd_name) >= 1 && cmd_name[0] == '#')
      {
	return 0;
      }
      
      if(strcmp(cmd_name, "cd") == 0)
      {
	char* path = cmd->argv[1] ? cmd->argv[1] : getenv("HOME");
	if (cd(path) < 0)
	{
	  char err_buf[max_length] = {0};
	  sprintf(err_buf, "cd: %s", path);
	  perror(err_buf);
	}

        continue;
      }
      if(strcmp(cmd_name, "exit") == 0)
      {
	int32_t status_code = cmd->argv[1] != NULL ? atoi(cmd->argv[1]) : 0;
	exit(status_code);
      }

      pids[i] = run_with_redir(cmd, n_pipes, pipes);
    }

    close_ALL_the_pipes(n_pipes, pipes);

    if (capture)
        close(capture_pipe[1]);

    int status;
    for (int i = 0; i < pl->n_cmds; ++i)
        waitpid(pids[i], &status, 0);

    if (capture)
    {
        char buffer[max_length];
        size_t total = 0;
        char* result = NULL;
        ssize_t n;

        while ((n = read(capture_pipe[0], buffer, sizeof(buffer))) > 0)
        {
            result = realloc(result, total + n + 1);
            memcpy(result + total, buffer, n);
            total += n;
        }

        if (!result)
            result = strdup("");

        result[total] = '\0';

        if (total > 0 && result[total-1] == '\n')
            result[total-1] = '\0';
	parser_allocated(result);
        *out = result;
        close(capture_pipe[0]);
    }

    return WEXITSTATUS(status);
}

int main(int argc, const char* argv[])
{
  FILE* stream = stdin;
  char* prompt = "arsh> ";
  const bool spawn_command = argc >= 2 && strcmp(argv[1], "-c") == 0;
  if(spawn_command)
  {
    if(argc !=3)
    {
      fprintf(stderr, "There must be exactly 3 arguments, when starting with a command '-c', but provided %d\n", argc);
      exit(1);
    }
    char* input = strdup(argv[2]);
    struct pipeline* pipeline = parse_pipeline(input);
    exit(execute_pipeline(pipeline, 0, NULL));
  }
  if(argc == 2)
  {
    FILE* file = fopen(argv[1], "r");
    if(file)
    {
      stream = file;
      prompt = NULL;
    }
  }

  struct sigaction s;
  s.sa_handler = sigint_handler;
  sigemptyset(&s.sa_mask);
  s.sa_flags = SA_RESTART;
  sigaction(SIGINT, &s, &s_old);

  int stat_loc;

  if(sigsetjmp(restart_jmp_buf, 1) == restart_value)
  {
    printf("\n");
  }
  jump_active = 1;
  
  while(true)
  {
    char* input = readline(prompt, stream);
    if(input == NULL) //CTRL + D
    {
      if(prompt)
      {
	printf("\nexit\n");
      }
      exit(0);
    }
    
    struct pipeline* pipeline = parse_pipeline(input);
    execute_pipeline(pipeline, 0, NULL);

    parser_free();
  }
}