[PATCH] POLL_OUT in ttys, misc bug fix

[PATCH] POLL_OUT in ttys, misc bug fix

[Sapan J . Bhatia]

[-- Attachment #1: Type: text/plain, Size: 719 bytes --]

Hello,

While fixing a flow control bug that truncated output to a terminal in
User Mode Linux, I noticed that the tty drivers only send POLL_IN on new
data being available but not POLL_OUT when the device is ready for new
data.

This patch fixes the bug in the line discipline and the pty driver.

Also, there's another minor bug in n_tty.c where write_chan returns
on a (retvalue < 0) unconditionally. This is a problem, since the type of
IO (BLOCKING / NON_BLOCKING) is stored in the tty, and if the console driver
returns a -EAGAIN (eg. in UML on getting an EAGAIN from the host kernel), 
write_chan returns even in the case of a blocking write, which is wrong
since the process doesn't expect it.

Regards,
Sapan


[-- Attachment #2: tty_patch.diff --]
[-- Type: text/plain, Size: 2466 bytes --]

--- /tmp/work/linux/drivers/char/pty.c	Fri Dec 21 23:11:54 2001
+++ /usr/src/linux-2.4.18/drivers/char/pty.c	Thu Mar 21 20:26:01 2002
@@ -5,6 +5,10 @@
  *
  *  Added support for a Unix98-style ptmx device.
  *    -- C. Scott Ananian <cananian@alumni.princeton.edu>, 14-Jan-1998
+ *  Added TTY_DO_WRITE_WAKEUP to enable n_tty to send POLL_OUT to
+ *      waiting writers -- Sapan Bhatia <sapan@corewars.org>
+ *
+ *
  */
 
 #include <linux/config.h>
@@ -331,6 +335,8 @@
 	clear_bit(TTY_OTHER_CLOSED, &tty->link->flags);
 	wake_up_interruptible(&pty->open_wait);
 	set_bit(TTY_THROTTLED, &tty->flags);
+	set_bit(TTY_DO_WRITE_WAKEUP, &tty->flags);
+
 	/*  Register a slave for the master  */
 	if (tty->driver.major == PTY_MASTER_MAJOR)
 		tty_register_devfs(&tty->link->driver,
--- /tmp/work/linux/drivers/char/n_tty.c	Fri Apr  6 23:12:55 2001
+++ /usr/src/linux-2.4.18/drivers/char/n_tty.c	Thu Mar 21 07:22:44 2002
@@ -23,6 +23,13 @@
  * 2000/01/20   Fixed SMP locking on put_tty_queue using bits of 
  *		the patch by Andrew J. Kroll <ag784@freenet.buffalo.edu>
  *		who actually finally proved there really was a race.
+ *
+ * 2002/03/18   Implemented n_tty_wakeup to send SIGIO POLL_OUTs to
+ *		waiting writing processes-Sapan Bhatia <sapan@corewars.org>
+ *
+ * 2002/03/19   Fixed write_chan to stay put if console driver returns
+ *              EAGAIN and not return since it returns an EAGAIN in a 
+ *		non-blocking operation-Sapan Bhatia <sapan@corewars.org>
  */
 
 #include <linux/types.h>
@@ -711,6 +718,22 @@
 	return 0;
 }
 
+/*
+ * Required for the ptys, serial driver etc. since processes
+ * that attach themselves to the master and rely on ASYNC
+ * IO must be woken up
+ */
+
+static void n_tty_write_wakeup(struct tty_struct *tty)
+{
+	if (tty->fasync)
+	{
+ 		set_bit(TTY_DO_WRITE_WAKEUP, &tty->flags);
+		kill_fasync(&tty->fasync, SIGIO, POLL_OUT);
+	}
+	return;
+}
+
 static void n_tty_receive_buf(struct tty_struct *tty, const unsigned char *cp,
 			      char *fp, int count)
 {
@@ -1156,7 +1179,7 @@
 		if (O_OPOST(tty) && !(test_bit(TTY_HW_COOK_OUT, &tty->flags))) {
 			while (nr > 0) {
 				ssize_t num = opost_block(tty, b, nr);
-				if (num < 0) {
+				if (num < 0 && num != -EAGAIN) {
 					retval = num;
 					goto break_out;
 				}
@@ -1236,6 +1259,6 @@
 	normal_poll,		/* poll */
 	n_tty_receive_buf,	/* receive_buf */
 	n_tty_receive_room,	/* receive_room */
-	0			/* write_wakeup */
+	n_tty_write_wakeup	/* write_wakeup */
 };
 

Re: [PATCH] POLL_OUT in ttys, misc bug fix

[Marcelo Tosatti]

Sapan, 

Have you actually tested the SIGIO POLL_OUT changes with an userspace app?

Thanks

On Wed, 27 Mar 2002, Sapan J . Bhatia wrote:

> Hello,
> 
> While fixing a flow control bug that truncated output to a terminal in
> User Mode Linux, I noticed that the tty drivers only send POLL_IN on new
> data being available but not POLL_OUT when the device is ready for new
> data.
> 
> This patch fixes the bug in the line discipline and the pty driver.
> 
> Also, there's another minor bug in n_tty.c where write_chan returns
> on a (retvalue < 0) unconditionally. This is a problem, since the type of
> IO (BLOCKING / NON_BLOCKING) is stored in the tty, and if the console driver
> returns a -EAGAIN (eg. in UML on getting an EAGAIN from the host kernel), 
> write_chan returns even in the case of a blocking write, which is wrong
> since the process doesn't expect it.
> 
> Regards,
> Sapan
> 
> 

Re: [PATCH] POLL_OUT in ttys, misc bug fix

[Sapan J . Bhatia]

    Hi Marcelo!

Yes, I have been testing them with userspace applications (most notably 
User Mode Linux).

The following program's output gets truncated on a pty running on a normal
kernel, but works fine when the patch is applied:
   www.corewars.org/pollout.c

Also, the version I'd sent you earlier did not have a fix for serial.c, which
exhibits the same problems. I've posted another version earlier today which 
includes this as well as a minor change to the second bug fix.
Please use the new version instead of the earlier one.

      Cheers,
    Sapan


On Thu, Apr 04, 2002 at 04:56:22PM -0300, Marcelo Tosatti wrote:
// 
// Sapan, 
// 
// Have you actually tested the SIGIO POLL_OUT changes with an userspace app?
// 
// Thanks
//