sock.c

LINUX 1.0 内核c源代码,是一份精简的linux内核源代码

  memcpy(fname, sockun.sun_path, sockaddr_len-UN_PATH_OFFSET);
  fname[sockaddr_len-UN_PATH_OFFSET] = '\0';
  old_fs = get_fs();
  set_fs(get_ds());
  i = open_namei(fname, 0, S_IFSOCK, &inode, NULL);
  set_fs(old_fs);
  if (i < 0) {
	dprintf(1, "UNIX: connect: can't open socket %s\n", fname);
	return(i);
  }
  serv_upd = unix_data_lookup(&sockun, sockaddr_len, inode);
  iput(inode);
  if (!serv_upd) {
	dprintf(1, "UNIX: connect: can't locate peer %s at inode 0x%x\n",
								fname, inode);
	return(-EINVAL);
  }
  if ((i = sock_awaitconn(sock, serv_upd->socket)) < 0) {
	dprintf(1, "UNIX: connect: can't await connection\n");
	return(i);
  }
  if (sock->conn) {
	unix_data_ref(UN_DATA(sock->conn));
	UN_DATA(sock)->peerupd = UN_DATA(sock->conn); /* ref server */
  }
  return(0);
}


/*
 * To do a socketpair, we just connect the two datas, easy!
 * Since we always wait on the socket inode, they're no contention
 * for a wait area, and deadlock prevention in the case of a process
 * writing to itself is, ignored, in true unix fashion!
 */
static int
unix_proto_socketpair(struct socket *sock1, struct socket *sock2)
{
  struct unix_proto_data *upd1 = UN_DATA(sock1), *upd2 = UN_DATA(sock2);

  unix_data_ref(upd1);
  unix_data_ref(upd2);
  upd1->peerupd = upd2;
  upd2->peerupd = upd1;
  return(0);
}


/* On accept, we ref the peer's data for safe writes. */
static int
unix_proto_accept(struct socket *sock, struct socket *newsock, int flags)
{
  struct socket *clientsock;

  dprintf(1, "UNIX: accept: socket 0x%x accepted via socket 0x%x\n",
							sock, newsock);

  /*
   * If there aren't any sockets awaiting connection,
   * then wait for one, unless nonblocking.
   */
  while(!(clientsock = sock->iconn)) {
	if (flags & O_NONBLOCK) return(-EAGAIN);
	interruptible_sleep_on(sock->wait);
	if (current->signal & ~current->blocked) {
		dprintf(1, "UNIX: accept: sleep was interrupted\n");
		return(-ERESTARTSYS);
	}
  }

  /*
   * Great. Finish the connection relative to server and client,
   * wake up the client and return the new fd to the server.
   */
  sock->iconn = clientsock->next;
  clientsock->next = NULL;
  newsock->conn = clientsock;
  clientsock->conn = newsock;
  clientsock->state = SS_CONNECTED;
  newsock->state = SS_CONNECTED;
  unix_data_ref(UN_DATA(clientsock));
  UN_DATA(newsock)->peerupd	       = UN_DATA(clientsock);
  UN_DATA(newsock)->sockaddr_un        = UN_DATA(sock)->sockaddr_un;
  UN_DATA(newsock)->sockaddr_len       = UN_DATA(sock)->sockaddr_len;
  wake_up_interruptible(clientsock->wait);
  return(0);
}


/* Gets the current name or the name of the connected socket. */
static int
unix_proto_getname(struct socket *sock, struct sockaddr *usockaddr,
		   int *usockaddr_len, int peer)
{
  struct unix_proto_data *upd;
  int len;
  int er;

  dprintf(1, "UNIX: getname: socket 0x%x for %s\n", sock, peer?"peer":"self");
  if (peer) {
	if (sock->state != SS_CONNECTED) {
		dprintf(1, "UNIX: getname: socket not connected\n");
		return(-EINVAL);
	}
	upd = UN_DATA(sock->conn);
  } else
	upd = UN_DATA(sock);

  er=verify_area(VERIFY_WRITE, usockaddr_len, sizeof(*usockaddr_len));
  if(er)
  	return er;
  if ((len = get_fs_long(usockaddr_len)) <= 0) return(-EINVAL);
  if (len > upd->sockaddr_len) len = upd->sockaddr_len;
  if (len) {
	er=verify_area(VERIFY_WRITE, usockaddr, len);
	if(er)
		return er;
	memcpy_tofs(usockaddr, &upd->sockaddr_un, len);
  }
  put_fs_long(len, usockaddr_len);
  return(0);
}


/* We read from our own buf. */
static int
unix_proto_read(struct socket *sock, char *ubuf, int size, int nonblock)
{
  struct unix_proto_data *upd;
  int todo, avail;
  int er;

  if ((todo = size) <= 0) return(0);
  upd = UN_DATA(sock);
  while(!(avail = UN_BUF_AVAIL(upd))) {
	if (sock->state != SS_CONNECTED) {
		dprintf(1, "UNIX: read: socket not connected\n");
		return((sock->state == SS_DISCONNECTING) ? 0 : -EINVAL);
	}
	dprintf(1, "UNIX: read: no data available...\n");
	if (nonblock) return(-EAGAIN);
	interruptible_sleep_on(sock->wait);
	if (current->signal & ~current->blocked) {
		dprintf(1, "UNIX: read: interrupted\n");
		return(-ERESTARTSYS);
	}
  }

  /*
   * Copy from the read buffer into the user's buffer,
   * watching for wraparound. Then we wake up the writer.
   */
   
  unix_lock(upd);
  do {
	int part, cando;

	if (avail <= 0) {
		printk("UNIX: read: AVAIL IS NEGATIVE!!!\n");
		send_sig(SIGKILL, current, 1);
		return(-EPIPE);
	}

	if ((cando = todo) > avail) cando = avail;
	if (cando >(part = BUF_SIZE - upd->bp_tail)) cando = part;
	dprintf(1, "UNIX: read: avail=%d, todo=%d, cando=%d\n",
	       					avail, todo, cando);
	if((er=verify_area(VERIFY_WRITE,ubuf,cando))<0)
	{
		unix_unlock(upd);
		return er;
	}
	memcpy_tofs(ubuf, upd->buf + upd->bp_tail, cando);
	upd->bp_tail =(upd->bp_tail + cando) &(BUF_SIZE-1);
	ubuf += cando;
	todo -= cando;
	if (sock->state == SS_CONNECTED)
		wake_up_interruptible(sock->conn->wait);
	avail = UN_BUF_AVAIL(upd);
  } while(todo && avail);
  unix_unlock(upd);
  return(size - todo);
}


/*
 * We write to our peer's buf. When we connected we ref'd this
 * peer so we are safe that the buffer remains, even after the
 * peer has disconnected, which we check other ways.
 */
static int
unix_proto_write(struct socket *sock, char *ubuf, int size, int nonblock)
{
  struct unix_proto_data *pupd;
  int todo, space;
  int er;

  if ((todo = size) <= 0) return(0);
  if (sock->state != SS_CONNECTED) {
	dprintf(1, "UNIX: write: socket not connected\n");
	if (sock->state == SS_DISCONNECTING) {
		send_sig(SIGPIPE, current, 1);
		return(-EPIPE);
	}
	return(-EINVAL);
  }
  pupd = UN_DATA(sock)->peerupd;	/* safer than sock->conn */

  while(!(space = UN_BUF_SPACE(pupd))) {
	dprintf(1, "UNIX: write: no space left...\n");
	if (nonblock) return(-EAGAIN);
	interruptible_sleep_on(sock->wait);
	if (current->signal & ~current->blocked) {
		dprintf(1, "UNIX: write: interrupted\n");
		return(-ERESTARTSYS);
	}
	if (sock->state == SS_DISCONNECTING) {
		dprintf(1, "UNIX: write: disconnected(SIGPIPE)\n");
		send_sig(SIGPIPE, current, 1);
		return(-EPIPE);
	}
  }

  /*
   * Copy from the user's buffer to the write buffer,
   * watching for wraparound. Then we wake up the reader.
   */
   
  unix_lock(pupd);
  
  do {
	int part, cando;

	if (space <= 0) {
		printk("UNIX: write: SPACE IS NEGATIVE!!!\n");
		send_sig(SIGKILL, current, 1);
		return(-EPIPE);
	}

	/*
	 * We may become disconnected inside this loop, so watch
	 * for it (peerupd is safe until we close).
	 */
	if (sock->state == SS_DISCONNECTING) {
		send_sig(SIGPIPE, current, 1);
		unix_unlock(pupd);
		return(-EPIPE);
	}
	if ((cando = todo) > space) cando = space;
	if (cando >(part = BUF_SIZE - pupd->bp_head)) cando = part;
	dprintf(1, "UNIX: write: space=%d, todo=%d, cando=%d\n",
	       					space, todo, cando);
	er=verify_area(VERIFY_READ, ubuf, cando);
	if(er)
	{
		unix_unlock(pupd);
		return er;
	}
	memcpy_fromfs(pupd->buf + pupd->bp_head, ubuf, cando);
	pupd->bp_head =(pupd->bp_head + cando) &(BUF_SIZE-1);
	ubuf += cando;
	todo -= cando;
	if (sock->state == SS_CONNECTED)
		wake_up_interruptible(sock->conn->wait);
	space = UN_BUF_SPACE(pupd);
  } while(todo && space);
  unix_unlock(pupd);
  return(size - todo);
}


static int
unix_proto_select(struct socket *sock, int sel_type, select_table * wait)
{
  struct unix_proto_data *upd, *peerupd;

  /* Handle server sockets specially. */
  if (sock->flags & SO_ACCEPTCON) {
	if (sel_type == SEL_IN) {
		dprintf(1, "UNIX: select: %sconnections pending\n",
		       				sock->iconn ? "" : "no ");
		if (sock->iconn) return(1);
		select_wait(sock->wait, wait);
		return(sock->iconn ? 1 : 0);
	}
	dprintf(1, "UNIX: select: nothing else for server socket\n");
	select_wait(sock->wait, wait);
	return(0);
  }

  if (sel_type == SEL_IN) {
	upd = UN_DATA(sock);
	dprintf(1, "UNIX: select: there is%s data available\n",
	       					UN_BUF_AVAIL(upd) ? "" : " no");
	if (UN_BUF_AVAIL(upd))	/* even if disconnected */
			return(1);
	else if (sock->state != SS_CONNECTED) {
		dprintf(1, "UNIX: select: socket not connected(read EOF)\n");
		return(1);
	}
	select_wait(sock->wait,wait);
	return(0);
  }
  if (sel_type == SEL_OUT) {
	if (sock->state != SS_CONNECTED) {
		dprintf(1, "UNIX: select: socket not connected(write EOF)\n");
		return(1);
	}
	peerupd = UN_DATA(sock->conn);
	dprintf(1, "UNIX: select: there is%s space available\n",
	       				UN_BUF_SPACE(peerupd) ? "" : " no");
	if (UN_BUF_SPACE(peerupd) > 0) return(1);
	select_wait(sock->wait,wait);
	return(0);
  }

  /* SEL_EX */
  dprintf(1, "UNIX: select: there are no exceptions here?!\n");
  return(0);
}


static int
unix_proto_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
{
  struct unix_proto_data *upd, *peerupd;
  int er;

  upd = UN_DATA(sock);
  peerupd = (sock->state == SS_CONNECTED) ? UN_DATA(sock->conn) : NULL;

  switch(cmd) {
	case TIOCINQ:
		if (sock->flags & SO_ACCEPTCON) return(-EINVAL);
		er=verify_area(VERIFY_WRITE,(void *)arg, sizeof(unsigned long));
		if(er)
			return er;
		if (UN_BUF_AVAIL(upd) || peerupd)
			put_fs_long(UN_BUF_AVAIL(upd),(unsigned long *)arg);
		  else
			put_fs_long(0,(unsigned long *)arg);
		break;
	case TIOCOUTQ:
		if (sock->flags & SO_ACCEPTCON) return(-EINVAL);
		er=verify_area(VERIFY_WRITE,(void *)arg, sizeof(unsigned long));
		if(er)
			return er;
		if (peerupd) put_fs_long(UN_BUF_SPACE(peerupd),
				   		(unsigned long *)arg);
		  else
			put_fs_long(0,(unsigned long *)arg);
		break;
	default:
		return(-EINVAL);
  }
  return(0);
}


static int
unix_open(struct inode * inode, struct file * file)
{
  int minor;

  dprintf(1, "UNIX: open\n");
  minor = MINOR(inode->i_rdev);
  if (minor != 0) return(-ENODEV);

  return(0);
}


static void
unix_close(struct inode * inode, struct file * file)
{
  dprintf(1, "UNIX: close\n");
}


static int
unix_ioctl(struct inode *inode, struct file *file,
	 unsigned int cmd, unsigned long arg)
{
  int minor, ret;
  int er;

  dprintf(1, "UNIX: ioctl(0x%X, 0x%X)\n", cmd, arg);
  minor = MINOR(inode->i_rdev);
  if (minor != 0) return(-ENODEV);

  ret = -EINVAL;
  switch(cmd) {
	case DDIOCSDBG:
		er=verify_area(VERIFY_READ,(void *)arg, sizeof(int));
		if(er)
			return er;
		unix_debug = get_fs_long((int *)arg);
		if (unix_debug != 0 && unix_debug != 1) {
			unix_debug = 0;
			return(-EINVAL);
		}
		return(0);
	case SIOCSIFLINK:
		printk("UNIX: cannot link streams!\n");
		break;
	default:
		break;
  }
  return(ret);
}




static struct file_operations unix_fops = {
  NULL,		/* LSEEK	*/
  NULL,		/* READ		*/
  NULL,		/* WRITE	*/
  NULL,		/* READDIR	*/
  NULL,		/* SELECT	*/
  unix_ioctl,	/* IOCTL	*/
  NULL,		/* MMAP		*/
  unix_open,	/* OPEN		*/
  unix_close	/* CLOSE	*/
};


static struct proto_ops unix_proto_ops = {
  AF_UNIX,
  unix_proto_create,
  unix_proto_dup,
  unix_proto_release,
  unix_proto_bind,
  unix_proto_connect,
  unix_proto_socketpair,
  unix_proto_accept,
  unix_proto_getname,
  unix_proto_read,
  unix_proto_write,
  unix_proto_select,
  unix_proto_ioctl,
  unix_proto_listen,
  unix_proto_send,
  unix_proto_recv,
  unix_proto_sendto,
  unix_proto_recvfrom,
  unix_proto_shutdown,
  unix_proto_setsockopt,
  unix_proto_getsockopt,
  NULL				/* unix_proto_fcntl	*/
};


void
unix_proto_init(struct ddi_proto *pro)
{
  struct unix_proto_data *upd;

  dprintf(1, "%s: init: initializing...\n", pro->name);
  if (register_chrdev(AF_UNIX_MAJOR, "af_unix", &unix_fops) < 0) {
	printk("%s: cannot register major device %d!\n",
						pro->name, AF_UNIX_MAJOR);
	return;
  }

  /* Tell SOCKET that we are alive... */
  (void) sock_register(unix_proto_ops.family, &unix_proto_ops);

  for(upd = unix_datas; upd <= last_unix_data; ++upd) {
	upd->refcnt = 0;
  }
}