tty: BKL pushdown

- Push the BKL down into the line disciplines
- Switch the tty layer to unlocked_ioctl
- Introduce a new ctrl_lock spin lock for the control bits
- Eliminate much of the lock_kernel use in n_tty
- Prepare to (but don't yet) call the drivers with the lock dropped
  on the paths that historically held the lock

BKL now primarily protects open/close/ldisc change in the tty layer

[jirislaby@gmail.com: a couple of fixes]
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/char/n_hdlc.c

@@ -578,27 +578,37 @@ static ssize_t n_hdlc_tty_read(struct tty_struct *tty, struct file *file,
 		return -EFAULT;
 	}
 
+	lock_kernel();
+
 	for (;;) {
-		if (test_bit(TTY_OTHER_CLOSED, &tty->flags))
+		if (test_bit(TTY_OTHER_CLOSED, &tty->flags)) {
+			unlock_kernel();
 			return -EIO;
+		}
 
 		n_hdlc = tty2n_hdlc (tty);
 		if (!n_hdlc || n_hdlc->magic != HDLC_MAGIC ||
-			 tty != n_hdlc->tty)
+			 tty != n_hdlc->tty) {
+			unlock_kernel();
 			return 0;
+		}
 
 		rbuf = n_hdlc_buf_get(&n_hdlc->rx_buf_list);
 		if (rbuf)
 			break;
 			
 		/* no data */
-		if (file->f_flags & O_NONBLOCK)
+		if (file->f_flags & O_NONBLOCK) {
+			unlock_kernel();
 			return -EAGAIN;
+		}
 			
 		interruptible_sleep_on (&tty->read_wait);
-		if (signal_pending(current))
+		if (signal_pending(current)) {
+			unlock_kernel();
 			return -EINTR;
 		}
+	}
 		
 	if (rbuf->count > nr)
 		/* frame too large for caller's buffer (discard frame) */
@@ -618,7 +628,7 @@ static ssize_t n_hdlc_tty_read(struct tty_struct *tty, struct file *file,
 		kfree(rbuf);
 	else	
 		n_hdlc_buf_put(&n_hdlc->rx_free_buf_list,rbuf);
-	
+	unlock_kernel();
 	return ret;
 	
 }	/* end of n_hdlc_tty_read() */
@@ -661,6 +671,8 @@ static ssize_t n_hdlc_tty_write(struct tty_struct *tty, struct file *file,
 		count = maxframe;
 	}
 	
+	lock_kernel();
+
 	add_wait_queue(&tty->write_wait, &wait);
 	set_current_state(TASK_INTERRUPTIBLE);
 	
@@ -695,7 +707,7 @@ static ssize_t n_hdlc_tty_write(struct tty_struct *tty, struct file *file,
 		n_hdlc_buf_put(&n_hdlc->tx_buf_list,tbuf);
 		n_hdlc_send_frames(n_hdlc,tty);
 	}
-
+	unlock_kernel();
 	return error;
 	
 }	/* end of n_hdlc_tty_write() */

drivers/char/n_r3964.c

@@ -1075,12 +1075,15 @@ static ssize_t r3964_read(struct tty_struct *tty, struct file *file,
 
 	TRACE_L("read()");
 
+	lock_kernel();
+
 	pClient = findClient(pInfo, task_pid(current));
 	if (pClient) {
 		pMsg = remove_msg(pInfo, pClient);
 		if (pMsg == NULL) {
 			/* no messages available. */
 			if (file->f_flags & O_NONBLOCK) {
+				unlock_kernel();
 				return -EAGAIN;
 			}
 			/* block until there is a message: */
@@ -1090,8 +1093,10 @@ static ssize_t r3964_read(struct tty_struct *tty, struct file *file,
 
 		/* If we still haven't got a message, we must have been signalled */
 
-		if (!pMsg)
+		if (!pMsg) {
+			unlock_kernel();
 			return -EINTR;
+		}
 
 		/* deliver msg to client process: */
 		theMsg.msg_id = pMsg->msg_id;
@@ -1102,12 +1107,15 @@ static ssize_t r3964_read(struct tty_struct *tty, struct file *file,
 		kfree(pMsg);
 		TRACE_M("r3964_read - msg kfree %p", pMsg);
 
-		if (copy_to_user(buf, &theMsg, count))
+		if (copy_to_user(buf, &theMsg, count)) {
+			unlock_kernel();
 			return -EFAULT;
+		}
 
 		TRACE_PS("read - return %d", count);
 		return count;
 	}
+	unlock_kernel();
 	return -EPERM;
 }
 
@@ -1156,6 +1164,8 @@ static ssize_t r3964_write(struct tty_struct *tty, struct file *file,
 	pHeader->locks = 0;
 	pHeader->owner = NULL;
 
+	lock_kernel();
+
 	pClient = findClient(pInfo, task_pid(current));
 	if (pClient) {
 		pHeader->owner = pClient;
@@ -1173,6 +1183,8 @@ static ssize_t r3964_write(struct tty_struct *tty, struct file *file,
 	add_tx_queue(pInfo, pHeader);
 	trigger_transmit(pInfo);
 
+	unlock_kernel();
+
 	return 0;
 }
 

drivers/char/n_tty.c

@@ -183,22 +183,24 @@ static void reset_buffer_flags(struct tty_struct *tty)
  *	at hangup) or when the N_TTY line discipline internally has to
  *	clean the pending queue (for example some signals).
  *
- *	FIXME: tty->ctrl_status is not spinlocked and relies on
- *	lock_kernel() still.
+ *	Locking: ctrl_lock
  */
 
 static void n_tty_flush_buffer(struct tty_struct *tty)
 {
+	unsigned long flags;
 	/* clear everything and unthrottle the driver */
 	reset_buffer_flags(tty);
 
 	if (!tty->link)
 		return;
 
+	spin_lock_irqsave(&tty->ctrl_lock, flags);
 	if (tty->link->packet) {
 		tty->ctrl_status |= TIOCPKT_FLUSHREAD;
 		wake_up_interruptible(&tty->link->read_wait);
 	}
+	spin_unlock_irqrestore(&tty->ctrl_lock, flags);
 }
 
 /**
@@ -264,7 +266,7 @@ static inline int is_continuation(unsigned char c, struct tty_struct *tty)
  *	relevant in the world today. If you ever need them, add them here.
  *
  *	Called from both the receive and transmit sides and can be called
- *	re-entrantly. Relies on lock_kernel() still.
+ *	re-entrantly. Relies on lock_kernel() for tty->column state.
  */
 
 static int opost(unsigned char c, struct tty_struct *tty)
@@ -275,6 +277,7 @@ static int opost(unsigned char c, struct tty_struct *tty)
 	if (!space)
 		return -1;
 
+	lock_kernel();
 	if (O_OPOST(tty)) {
 		switch (c) {
 		case '\n':
@@ -323,6 +326,7 @@ static int opost(unsigned char c, struct tty_struct *tty)
 		}
 	}
 	tty->driver->put_char(tty, c);
+	unlock_kernel();
 	return 0;
 }
 
@@ -337,7 +341,8 @@ static int opost(unsigned char c, struct tty_struct *tty)
  *	the simple cases normally found and helps to generate blocks of
  *	symbols for the console driver and thus improve performance.
  *
- *	Called from write_chan under the tty layer write lock.
+ *	Called from write_chan under the tty layer write lock. Relies
+ *	on lock_kernel for the tty->column state.
  */
 
 static ssize_t opost_block(struct tty_struct *tty,
@@ -353,6 +358,7 @@ static ssize_t opost_block(struct tty_struct *tty,
 	if (nr > space)
 		nr = space;
 
+	lock_kernel();
 	for (i = 0, cp = buf; i < nr; i++, cp++) {
 		switch (*cp) {
 		case '\n':
@@ -387,6 +393,7 @@ static ssize_t opost_block(struct tty_struct *tty,
 	if (tty->driver->flush_chars)
 		tty->driver->flush_chars(tty);
 	i = tty->driver->write(tty, buf, i);
+	unlock_kernel();
 	return i;
 }
 
@@ -1194,6 +1201,11 @@ extern ssize_t redirected_tty_write(struct file *, const char __user *,
  *	Perform job control management checks on this file/tty descriptor
  *	and if appropriate send any needed signals and return a negative
  *	error code if action should be taken.
+ *
+ *	FIXME:
+ *	Locking: None - redirected write test is safe, testing
+ *	current->signal should possibly lock current->sighand
+ *	pgrp locking ?
  */
 
 static int job_control(struct tty_struct *tty, struct file *file)
@@ -1246,6 +1258,7 @@ static ssize_t read_chan(struct tty_struct *tty, struct file *file,
 	ssize_t size;
 	long timeout;
 	unsigned long flags;
+	int packet;
 
 do_it_again:
 
@@ -1289,16 +1302,19 @@ static ssize_t read_chan(struct tty_struct *tty, struct file *file,
 		if (mutex_lock_interruptible(&tty->atomic_read_lock))
 			return -ERESTARTSYS;
 	}
+	packet = tty->packet;
 
 	add_wait_queue(&tty->read_wait, &wait);
 	while (nr) {
 		/* First test for status change. */
-		if (tty->packet && tty->link->ctrl_status) {
+		if (packet && tty->link->ctrl_status) {
 			unsigned char cs;
 			if (b != buf)
 				break;
+			spin_lock_irqsave(&tty->link->ctrl_lock, flags);
 			cs = tty->link->ctrl_status;
 			tty->link->ctrl_status = 0;
+			spin_unlock_irqrestore(&tty->link->ctrl_lock, flags);
 			if (tty_put_user(tty, cs, b++)) {
 				retval = -EFAULT;
 				b--;
@@ -1333,6 +1349,7 @@ static ssize_t read_chan(struct tty_struct *tty, struct file *file,
 				retval = -ERESTARTSYS;
 				break;
 			}
+			/* FIXME: does n_tty_set_room need locking ? */
 			n_tty_set_room(tty);
 			timeout = schedule_timeout(timeout);
 			continue;
@@ -1340,7 +1357,7 @@ static ssize_t read_chan(struct tty_struct *tty, struct file *file,
 		__set_current_state(TASK_RUNNING);
 
 		/* Deal with packet mode. */
-		if (tty->packet && b == buf) {
+		if (packet && b == buf) {
 			if (tty_put_user(tty, TIOCPKT_DATA, b++)) {
 				retval = -EFAULT;
 				b--;
@@ -1388,6 +1405,8 @@ static ssize_t read_chan(struct tty_struct *tty, struct file *file,
 				break;
 		} else {
 			int uncopied;
+			/* The copy function takes the read lock and handles
+			   locking internally for this case */
 			uncopied = copy_from_read_buf(tty, &b, &nr);
 			uncopied += copy_from_read_buf(tty, &b, &nr);
 			if (uncopied) {
@@ -1429,7 +1448,6 @@ static ssize_t read_chan(struct tty_struct *tty, struct file *file,
 		 goto do_it_again;
 
 	n_tty_set_room(tty);
-
 	return retval;
 }
 

drivers/char/pty.c

@@ -181,6 +181,7 @@ static int pty_set_lock(struct tty_struct *tty, int __user * arg)
 static void pty_flush_buffer(struct tty_struct *tty)
 {
 	struct tty_struct *to = tty->link;
+	unsigned long flags;
 	
 	if (!to)
 		return;
@@ -189,8 +190,10 @@ static void pty_flush_buffer(struct tty_struct *tty)
 		to->ldisc.flush_buffer(to);
 	
 	if (to->packet) {
+		spin_lock_irqsave(&tty->ctrl_lock, flags);
 		tty->ctrl_status |= TIOCPKT_FLUSHWRITE;
 		wake_up_interruptible(&to->read_wait);
+		spin_unlock_irqrestore(&tty->ctrl_lock, flags);
 	}
 }
 

drivers/char/tty_ioctl.c

@@ -395,6 +395,7 @@ static void change_termios(struct tty_struct *tty, struct ktermios *new_termios)
 	int canon_change;
 	struct ktermios old_termios = *tty->termios;
 	struct tty_ldisc *ld;
+	unsigned long flags;
 
 	/*
 	 *	Perform the actual termios internal changes under lock.
@@ -429,11 +430,13 @@ static void change_termios(struct tty_struct *tty, struct ktermios *new_termios)
 				STOP_CHAR(tty) == '\023' &&
 				START_CHAR(tty) == '\021');
 		if (old_flow != new_flow) {
+			spin_lock_irqsave(&tty->ctrl_lock, flags);
 			tty->ctrl_status &= ~(TIOCPKT_DOSTOP | TIOCPKT_NOSTOP);
 			if (new_flow)
 				tty->ctrl_status |= TIOCPKT_DOSTOP;
 			else
 				tty->ctrl_status |= TIOCPKT_NOSTOP;
+			spin_unlock_irqrestore(&tty->ctrl_lock, flags);
 			wake_up_interruptible(&tty->link->read_wait);
 		}
 	}
@@ -905,6 +908,7 @@ int n_tty_ioctl(struct tty_struct *tty, struct file *file,
 		       unsigned int cmd, unsigned long arg)
 {
 	struct tty_struct *real_tty;
+	unsigned long flags;
 	int retval;
 
 	if (tty->driver->type == TTY_DRIVER_TYPE_PTY &&
@@ -963,6 +967,7 @@ int n_tty_ioctl(struct tty_struct *tty, struct file *file,
 			return -ENOTTY;
 		if (get_user(pktmode, (int __user *) arg))
 			return -EFAULT;
+		spin_lock_irqsave(&tty->ctrl_lock, flags);
 		if (pktmode) {
 			if (!tty->packet) {
 				tty->packet = 1;
@@ -970,6 +975,7 @@ int n_tty_ioctl(struct tty_struct *tty, struct file *file,
 			}
 		} else
 			tty->packet = 0;
+		spin_unlock_irqrestore(&tty->ctrl_lock, flags);
 		return 0;
 	}
 	default:

drivers/char/vt.c

@@ -2541,6 +2541,9 @@ int tioclinux(struct tty_struct *tty, unsigned long arg)
 	if (get_user(type, p))
 		return -EFAULT;
 	ret = 0;
+
+	lock_kernel();
+
 	switch (type)
 	{
 		case TIOCL_SETSEL:
@@ -2615,6 +2618,7 @@ int tioclinux(struct tty_struct *tty, unsigned long arg)
 			ret = -EINVAL;
 			break;
 	}
+	unlock_kernel();
 	return ret;
 }
 

fs/compat_ioctl.c

@@ -1046,7 +1046,7 @@ static int vt_check(struct file *file)
 	struct inode *inode = file->f_path.dentry->d_inode;
 	struct vc_data *vc;
 	
-	if (file->f_op->ioctl != tty_ioctl)
+	if (file->f_op->unlocked_ioctl != tty_ioctl)
 		return -EINVAL;
 	                
 	tty = (struct tty_struct *)file->private_data;

include/linux/tty.h

@@ -183,6 +183,7 @@ struct tty_struct {
 	int index;
 	struct tty_ldisc ldisc;
 	struct mutex termios_mutex;
+	spinlock_t ctrl_lock;
 	struct ktermios *termios, *termios_locked;
 	char name[64];
 	struct pid *pgrp;
@@ -323,8 +324,7 @@ extern void tty_ldisc_put(int);
 extern void tty_wakeup(struct tty_struct *tty);
 extern void tty_ldisc_flush(struct tty_struct *tty);
 
-extern int tty_ioctl(struct inode *inode, struct file *file, unsigned int cmd,
-		     unsigned long arg);
+extern long tty_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 extern int tty_mode_ioctl(struct tty_struct *tty, struct file *file,
 			unsigned int cmd, unsigned long arg);
 extern int tty_perform_flush(struct tty_struct *tty, unsigned long arg);