Re: TTY loosing data with u_serial gadget

[Stefan Bigler]

Hi Greg

Thanks for your quick answer. Meanwhile we ported our board to 2.6.38.
We recognize the same problem again.
I had also a look at the relevant fixes, a lot is done but I could not find
the required protection of the attribute receive_room.

I added a printk in case of failure were receive_room shows more
space that is really available in the queue and also used.

The printk is attached below and also the log.

Regards
Stefan


Log:
flush_to_ldisc, receive_room to small count=201, receive_room=287, 
real_room=0
flush_to_ldisc, receive_room to small count=225, receive_room=1031, 
real_room=0
flush_to_ldisc, receive_room to small count=33, receive_room=4095, 
real_room=0


diff --git a/drivers/tty/tty_buffer.c b/drivers/tty/tty_buffer.c
index d8210ca..fa9ca62 100644
--- a/drivers/tty/tty_buffer.c
+++ b/drivers/tty/tty_buffer.c
@@ -442,6 +442,21 @@ static void flush_to_ldisc(struct work_struct *work)
                line discipline as we want to empty the queue */
             if (test_bit(TTY_FLUSHPENDING, &tty->flags))
                 break;
+
+            /* add printk in case of inconsistante size */
+            if (tty->raw) {
+                unsigned long flags_r;
+                int real_room;
+                spin_lock_irqsave(&tty->read_lock, flags_r);
+                real_room = N_TTY_BUF_SIZE -tty->read_cnt - 1;
+                if (real_room < min(count, tty->receive_room)) {
+                    printk(KERN_ERR "flush_to_ldisc, receive_room to 
small count=%d,"
+                        " receive_room=%d, real_room=%d\n",
+                        count, tty->receive_room, real_room);
+                }
+                spin_unlock_irqrestore(&tty->read_lock,    flags_r);
+            }
+
             if (!tty->receive_room || seen_tail) {
                 schedule_delayed_work(&tty->buf.work, 1);
                 break;





Am 03/17/2011 01:04 AM, wrote Greg KH:
> On Wed, Mar 16, 2011 at 09:47:50PM +0100, Stefan Bigler wrote:
> > Hi all
> >
> > I'm facing a problem with TTY loosing data using u_serial gadget.
> >
> > I have a MPC8247 based system running 2.6.33.
> Oh, that's an old kernel, lots of work on the tty layer has happened
> since then that would fix issues like this.
>
> If you are stuck using this kernel, please go ask your vendor for
> support for it.
>
> If not, can you try out 2.6.38 and let us, and the
> linux-usb@xxxxxxxxxxxxxxx list know if you still have problems with it?
>
> thanks,
>
> greg k-h

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/