Hello,
I make driver for Point of Sale Printer, a wide range of Printer use

only a DTR/DSR hardware-handshaking. When I use a handshaking in the

userspace, the Printr has a overrun problem and our customer has a

problem with the tax office.
my Patch relaize a simple DTR/DSR handhake with a small change of the

code. The change auf the userspace is very simple. e.g.
cflags |= CDTRDSR;
The change of the stty tool at 2 lines.
Michael
diff -Nur linux-2.6/drivers/serial/8250.c linux-2.6.dsr_test/drivers/serial/8250.c

--- linux-2.6/drivers/serial/8250.c	2007-08-28 11:31:48.000000000 +0200

+++ linux-2.6.dsr_test/drivers/serial/8250.c	2007-08-31 13:20:57.000000000 +0200

@@ -1400,12 +1400,15 @@

 	    up->port.info != NULL) {

 		if (status & UART_MSR_TERI)

 			up->port.icount.rng++;

-		if (status & UART_MSR_DDSR)

-			up->port.icount.dsr++;

 		if (status & UART_MSR_DDCD)

 			uart_handle_dcd_change(&up->port, status & UART_MSR_DCD);

-		if (status & UART_MSR_DCTS)

-			uart_handle_cts_change(&up->port, status & UART_MSR_CTS);

+		if (status & (UART_MSR_DCTS|UART_MSR_DDSR)) {

+			if (status & UART_MSR_DDSR)

+				up->port.icount.dsr++;

+			else

+				up->port.icount.cts++;

+			uart_handle_dsr_cts_change(&up->port, status & UART_MSR_CTS, status & UART_MSR_DSR);

+		}
 		wake_up_interruptible(&up->port.info->delta_msr_wait);

 	}

diff -Nur linux-2.6/drivers/serial/serial_core.c linux-2.6.dsr_test/drivers/serial/serial_core.c

--- linux-2.6/drivers/serial/serial_core.c	2007-08-28 11:31:48.000000000 +0200

+++ linux-2.6.dsr_test/drivers/serial/serial_core.c	2007-08-31 13:29:22.000000000 +0200

@@ -191,6 +191,13 @@

 				info->tty->hw_stopped = 1;

 			spin_unlock_irq(&port->lock);

 		}

+

+		if (info->flags & UIF_DSR_FLOW) {

+			spin_lock_irq(&port->lock);

+			if (!(port->ops->get_mctrl(port) & TIOCM_DSR))

+				info->tty->hw_stopped = 1;

+			spin_unlo...
[ message continues ]

On Fri, 31 Aug 2007 15:17:41 +0200
googling for CDTRDSR is useful.  I see from

http://lkml.org/lkml/2006/12/10/49 that Russell was generally supportive,
This will break all architectures except for i386.  So for now I guess we

can do this: 
--- a/include/linux/termios.h~serial-8250-implement-dtr-dsr-handshaking-fix

+++ a/include/linux/termios.h

@@ -4,4 +4,8 @@

 #include <linux/types.h>

 #include <asm/termios.h>
+#ifndef CDTRDSR

+#define CDTRDSR 0 /* remove this when all architectures have a definition */

+#endif

+

 #endif

_
but that's temporary.  If we decide to proceed with this change then we

should add CDTRDSR to each arch's termbits.h.  That'll be pretty

straightforward to arrange.
Actually, let's spam them ...
--- a/include/linux/termios.h~serial-8250-implement-dtr-dsr-handshaking-fix

+++ a/include/linux/termios.h

@@ -4,4 +4,9 @@

 #include <linux/types.h>

 #include <asm/termios.h>
+#ifndef CDTRDSR

+#warning This architecture should implement CDTRDSR

+#define CDTRDSR 0 /* remove this when all architectures have a definition */

+#endif

+

 #endif

_
-

Partially implemented - my point 1 seems to have been addressed but not

point 2.  Point 2 is rather important, otherwise you can end up with a

port in stopped mode with no way to get it transmitting.
Also, it doesn't seem to be tweaking the DTR signal at all, so how does

it tell the remote end to stop transmitting?
Essentially, the author needs to search for CRTSCTS in serial_core.c and

*carefully* consider whether they need to modify those places as well.

I stress carefully, because at the moment we start the tty whenever

we move from CRTSCTS to no CRTSCTS.  If you're moving CRTSCTS->CDTRDSR

then you only want to do this iff DSR is asserted.
--

Russell King

 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/

 maintainer of:

-