]> pilppa.com Git - linux-2.6-omap-h63xx.git/commitdiff
[PATCH] tty: Fix bits and note more bits to fix
authorAlan Cox <alan@lxorguk.ukuu.org.uk>
Fri, 29 Sep 2006 09:00:58 +0000 (02:00 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Fri, 29 Sep 2006 16:18:18 +0000 (09:18 -0700)
If your driver implements "break on" and "break off" this ensures you won't
get multiple overlapping requests or requests in parallel.  If your driver
has its own break handling then its still your problem as the driver
author.

Break is also now serialized against writes from user space properly but no
new guarantees are made driver level about writes from the line discipline
itself (eg flow control or echo)

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
drivers/char/tty_io.c

index 48b59006afcd341a8a7d2762e16bfd8b027b8203..63543731977846822c1270d36b0c3ca5547f52f7 100644 (file)
@@ -2719,6 +2719,8 @@ static int tty_fasync(int fd, struct file * filp, int on)
  *     Locking:
  *             Called functions take tty_ldisc_lock
  *             current->signal->tty check is safe without locks
+ *
+ *     FIXME: may race normal receive processing
  */
 
 static int tiocsti(struct tty_struct *tty, char __user *p)
@@ -2878,9 +2880,7 @@ static int fionbio(struct file *file, int __user *p)
  *     Locking:
  *             Takes tasklist lock internally to walk sessions
  *             Takes task_lock() when updating signal->tty
- *
- *     FIXME: tty_mutex is needed to protect signal->tty references.
- *     FIXME: why task_lock on the signal->tty reference ??
+ *             Takes tty_mutex() to protect tty instance
  *
  */
 
@@ -2915,9 +2915,11 @@ static int tiocsctty(struct tty_struct *tty, int arg)
                } else
                        return -EPERM;
        }
+       mutex_lock(&tty_mutex);
        task_lock(current);
        current->signal->tty = tty;
        task_unlock(current);
+       mutex_unlock(&tty_mutex);
        current->signal->tty_old_pgrp = 0;
        tty->session = current->signal->session;
        tty->pgrp = process_group(current);
@@ -2957,8 +2959,6 @@ static int tiocgpgrp(struct tty_struct *tty, struct tty_struct *real_tty, pid_t
  *     permitted where the tty session is our session.
  *
  *     Locking: None
- *
- *     FIXME: current->signal->tty referencing is unsafe.
  */
 
 static int tiocspgrp(struct tty_struct *tty, struct tty_struct *real_tty, pid_t __user *p)
@@ -3037,19 +3037,20 @@ static int tiocsetd(struct tty_struct *tty, int __user *p)
  *     timed break functionality.
  *
  *     Locking:
- *             None
+ *             atomic_write_lock serializes
  *
- *     FIXME:
- *             What if two overlap
  */
 
 static int send_break(struct tty_struct *tty, unsigned int duration)
 {
+       if (mutex_lock_interruptible(&tty->atomic_write_lock))
+               return -EINTR;
        tty->driver->break_ctl(tty, -1);
        if (!signal_pending(current)) {
                msleep_interruptible(duration);
        }
        tty->driver->break_ctl(tty, 0);
+       mutex_unlock(&tty->atomic_write_lock);
        if (signal_pending(current))
                return -EINTR;
        return 0;
@@ -3142,6 +3143,8 @@ int tty_ioctl(struct inode * inode, struct file * file,
        if (tty_paranoia_check(tty, inode, "tty_ioctl"))
                return -EINVAL;
 
+       /* CHECKME: is this safe as one end closes ? */
+
        real_tty = tty;
        if (tty->driver->type == TTY_DRIVER_TYPE_PTY &&
            tty->driver->subtype == PTY_TYPE_MASTER)