Fix staircase effect on Altix serial console.

Fix staircase effect on Altix serial console.

[Andreas Schwab]

Serial console output on Altix didn't convert NL to CRNL.

Andreas.

--- linux-2.6.3/drivers/char/sn_serial.c.~1~	2004-03-02 16:41:09.000000000 +0100
+++ linux-2.6.3/drivers/char/sn_serial.c	2004-03-02 22:51:09.049180364 +0100
@@ -984,6 +984,7 @@ static void
 sn_sal_console_write(struct console *co, const char *s, unsigned count)
 {
 	unsigned long flags;
+	const char *s1;
 
 	BUG_ON(!sn_sal_is_asynch);
 
@@ -991,15 +992,36 @@ sn_sal_console_write(struct console *co,
 	 * oops, kdb, panic, etc.  make sure they get it. */
 	if (spin_is_locked(&sn_sal_lock)) {
 		synch_flush_xmit();
+		/* Output '\r' before each '\n' */
+		while ((s1 = memchr(s, '\n', count)) != NULL) {
+			sn_func->sal_puts(s, s1 - s);
+			sn_func->sal_puts("\r\n", 2);
+			count -= s1 + 1 - s;
+			s = s1 + 1;
+		}
 		sn_func->sal_puts(s, count);
 	}
 	else if (in_interrupt()) {
 		spin_lock_irqsave(&sn_sal_lock, flags);
 		synch_flush_xmit();
 		spin_unlock_irqrestore(&sn_sal_lock, flags);
+		/* Output '\r' before each '\n' */
+		while ((s1 = memchr(s, '\n', count)) != NULL) {
+			sn_func->sal_puts(s, s1 - s);
+			sn_func->sal_puts("\r\n", 2);
+			count -= s1 + 1 - s;
+			s = s1 + 1;
+		}
 		sn_func->sal_puts(s, count);
 	}
 	else
+		/* Output '\r' before each '\n' */
+		while ((s1 = memchr(s, '\n', count)) != NULL) {
+			sn_sal_write(NULL, 0, s, s1 - s);
+			sn_sal_write(NULL, 0, "\r\n", 2);
+			count -= s1 + 1 - s;
+			s = s1 + 1;
+		}
 		sn_sal_write(NULL, 0, s, count);
 }
 

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Fix staircase effect on Altix serial console.

[David Mosberger]

>>>>> On Wed, 03 Mar 2004 00:49:20 +0100, Andreas Schwab <schwab@suse.de> said:

  Andreas> Serial console output on Altix didn't convert NL to CRNL.

Somebody from SGI will have to review/OK this.

	--david

Re: Fix staircase effect on Altix serial console.

[Jesse Barnes]

On Tue, Mar 02, 2004 at 06:56:55PM -0800, David Mosberger wrote:
> >>>>> On Wed, 03 Mar 2004 00:49:20 +0100, Andreas Schwab <schwab@suse.de> said:
> 
>   Andreas> Serial console output on Altix didn't convert NL to CRNL.
> 
> Somebody from SGI will have to review/OK this.

Andreas, on which platform and configuration did you see this problem?
It appears that depending on your console configuration, the system
controller will eat/convert characters.  I'm trying to get it fixed
properly there...

Jesse

Re: Fix staircase effect on Altix serial console.

[Andreas Schwab]

jbarnes@sgi.com (Jesse Barnes) writes:

> On Tue, Mar 02, 2004 at 06:56:55PM -0800, David Mosberger wrote:
>> >>>>> On Wed, 03 Mar 2004 00:49:20 +0100, Andreas Schwab <schwab@suse.de> said:
>> 
>>   Andreas> Serial console output on Altix didn't convert NL to CRNL.
>> 
>> Somebody from SGI will have to review/OK this.
>
> Andreas, on which platform and configuration did you see this problem?

It's on the 8P Altix (can't remember the exact platform name).

> It appears that depending on your console configuration, the system
> controller will eat/convert characters.  I'm trying to get it fixed
> properly there...

The system controller should not change the output in any way.  It's the
job of the serial console write function (which is used for printk) to
transform NL to CRNL, which is what I have implemented here.  Note that
normal serial output (which goes through the tty layer) does not exhibit
this problem.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Fix staircase effect on Altix serial console.

[Robin Holt]

On Wed, Mar 03, 2004 at 11:57:21AM +0100, Andreas Schwab wrote:
> jbarnes@sgi.com (Jesse Barnes) writes:
> 
> > On Tue, Mar 02, 2004 at 06:56:55PM -0800, David Mosberger wrote:
> >> >>>>> On Wed, 03 Mar 2004 00:49:20 +0100, Andreas Schwab <schwab@suse.de> said:
> >> 
> >>   Andreas> Serial console output on Altix didn't convert NL to CRNL.
> >> 
> >> Somebody from SGI will have to review/OK this.
> >
> > Andreas, on which platform and configuration did you see this problem?
> 
> It's on the 8P Altix (can't remember the exact platform name).
> 

Are you seeing the missing carriage return on output at the L1 or
on a terminal connected to the serial port?

> > It appears that depending on your console configuration, the system
> > controller will eat/convert characters.  I'm trying to get it fixed
> > properly there...
> 
> The system controller should not change the output in any way.  It's the
> job of the serial console write function (which is used for printk) to
> transform NL to CRNL, which is what I have implemented here.  Note that
> normal serial output (which goes through the tty layer) does not exhibit
> this problem.
> 
> Andreas.
> 
> -- 
> Andreas Schwab, SuSE Labs, schwab@suse.de
> SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
> Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
> "And now for something completely different."
> -
> To unsubscribe from this list: send the line "unsubscribe linux-ia64" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Fix staircase effect on Altix serial console.

[Andreas Schwab]

Robin Holt <holt@sgi.com> writes:

> On Wed, Mar 03, 2004 at 11:57:21AM +0100, Andreas Schwab wrote:
>> jbarnes@sgi.com (Jesse Barnes) writes:
>> 
>> > On Tue, Mar 02, 2004 at 06:56:55PM -0800, David Mosberger wrote:
>> >> >>>>> On Wed, 03 Mar 2004 00:49:20 +0100, Andreas Schwab <schwab@suse.de> said:
>> >> 
>> >>   Andreas> Serial console output on Altix didn't convert NL to CRNL.
>> >> 
>> >> Somebody from SGI will have to review/OK this.
>> >
>> > Andreas, on which platform and configuration did you see this problem?
>> 
>> It's on the 8P Altix (can't remember the exact platform name).
>> 
>
> Are you seeing the missing carriage return on output at the L1 or
> on a terminal connected to the serial port?

I don't know how to differenciate between these two things.  The serial
terminal is connected to the L1 system controller.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Fix staircase effect on Altix serial console.

[Robin Holt]

On Wed, Mar 03, 2004 at 02:34:33PM +0100, Andreas Schwab wrote:
> Robin Holt <holt@sgi.com> writes:
> 
> > On Wed, Mar 03, 2004 at 11:57:21AM +0100, Andreas Schwab wrote:
> >> jbarnes@sgi.com (Jesse Barnes) writes:
> >> 
> >> > On Tue, Mar 02, 2004 at 06:56:55PM -0800, David Mosberger wrote:
> >> >> >>>>> On Wed, 03 Mar 2004 00:49:20 +0100, Andreas Schwab <schwab@suse.de> said:
> >> >> 
> >> >>   Andreas> Serial console output on Altix didn't convert NL to CRNL.
> >> >> 
> >> >> Somebody from SGI will have to review/OK this.
> >> >
> >> > Andreas, on which platform and configuration did you see this problem?
> >> 
> >> It's on the 8P Altix (can't remember the exact platform name).
> >> 
> >
> > Are you seeing the missing carriage return on output at the L1 or
> > on a terminal connected to the serial port?
> 
> I don't know how to differenciate between these two things.  The serial
> terminal is connected to the L1 system controller.

Are you connected via usb or a DB-9 serial connector?

Robin

Re: Fix staircase effect on Altix serial console.

[Andreas Schwab]

Robin Holt <holt@sgi.com> writes:

> Are you connected via usb or a DB-9 serial connector?

DB-9, I think.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Fix staircase effect on Altix serial console.

[Robin Holt]

On Wed, Mar 03, 2004 at 03:20:46PM +0100, Andreas Schwab wrote:
> Robin Holt <holt@sgi.com> writes:
> 
> > Are you connected via usb or a DB-9 serial connector?
> 
> DB-9, I think.

The last time I connected to the DB-9 on the back of an Altix,
I was using minicom (I think) and just changed it so it
would interpret LF as CRLF.  I seem to remember something
like LF <--> CRLF on the line.  Don't remember if that was
exactly what it looked like.

Is this something that should really be done by the terminal
and not the serial driver?  I don't have any objections
either way, but I would guess we would want to stay as
close to whatever x86 linux is doing.  I plead ignorance at
this point.

Robin

Re: Fix staircase effect on Altix serial console.

[Andreas Schwab]

Robin Holt <holt@sgi.com> writes:

> On Wed, Mar 03, 2004 at 03:20:46PM +0100, Andreas Schwab wrote:
>> Robin Holt <holt@sgi.com> writes:
>> 
>> > Are you connected via usb or a DB-9 serial connector?
>> 
>> DB-9, I think.
>
> The last time I connected to the DB-9 on the back of an Altix,
> I was using minicom (I think) and just changed it so it
> would interpret LF as CRLF.

This is plain wrong.  Just try running a fullscreen program like vi and
watch your screen being garbled.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Fix staircase effect on Altix serial console.

[Jesse Barnes]

On Wed, Mar 03, 2004 at 11:57:21AM +0100, Andreas Schwab wrote:
> The system controller should not change the output in any way.  It's the
> job of the serial console write function (which is used for printk) to

Right, but I think our L2 system controller *does* molest the output
(and the 2.6 console driver relies on this behavior).  I've filed a bug
with our system controller group to make the behavior of the L2 and L1
system controllers consistent (I believe you're using an L1), which I
think is the proper fix.

Thanks,
Jesse

Re: Fix staircase effect on Altix serial console.

[Jesse Barnes]

On Wed, Mar 03, 2004 at 09:19:37AM -0800, Jesse Barnes wrote:
> On Wed, Mar 03, 2004 at 11:57:21AM +0100, Andreas Schwab wrote:
> > The system controller should not change the output in any way.  It's the
> > job of the serial console write function (which is used for printk) to
> 
> Right, but I think our L2 system controller *does* molest the output
> (and the 2.6 console driver relies on this behavior).  I've filed a bug
> with our system controller group to make the behavior of the L2 and L1
> system controllers consistent (I believe you're using an L1), which I
> think is the proper fix.

In case I wasn't clear: I think your patch would be ok for systems that
use L1s, but would result in extra CRs for systems using L2s as their
console, so we really need to fix the system controllers instead.

Thanks,
Jesse

Re: Fix staircase effect on Altix serial console.

[Andreas Schwab]

jbarnes@sgi.com (Jesse Barnes) writes:

> On Wed, Mar 03, 2004 at 11:57:21AM +0100, Andreas Schwab wrote:
>> The system controller should not change the output in any way.  It's the
>> job of the serial console write function (which is used for printk) to
>
> Right, but I think our L2 system controller *does* molest the output
> (and the 2.6 console driver relies on this behavior).

This is a bug.

> I've filed a bug with our system controller group to make the behavior
> of the L2 and L1 system controllers consistent (I believe you're using
> an L1), which I think is the proper fix.

No, it is the *wrong* thing to do.  If you do stty -onlcr you should get
just one NL, nothing else.  It's the job of the tty layer and the serial
console write function to translate NL to CRNL, nobody else should mangle
the output in any way.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Fix staircase effect on Altix serial console.

[Jesse Barnes]

On Wed, Mar 03, 2004 at 06:36:21PM +0100, Andreas Schwab wrote:
> > I've filed a bug with our system controller group to make the behavior
> > of the L2 and L1 system controllers consistent (I believe you're using
> > an L1), which I think is the proper fix.
> 
> No, it is the *wrong* thing to do.  If you do stty -onlcr you should get
> just one NL, nothing else.  It's the job of the tty layer and the serial
> console write function to translate NL to CRNL, nobody else should mangle
> the output in any way.

Right, that's why I'm saying that the l2 shouldn't mangle output, nor
should the PROM.  They should just pass through whatever the serial
driver hands them.  I think we're in agreement here?

Jesse

Re: Fix staircase effect on Altix serial console.

[Jesse Barnes]

On Wed, Mar 03, 2004 at 12:44:49PM -0800, Jesse Barnes wrote:
> On Wed, Mar 03, 2004 at 06:36:21PM +0100, Andreas Schwab wrote:
> > > I've filed a bug with our system controller group to make the behavior
> > > of the L2 and L1 system controllers consistent (I believe you're using
> > > an L1), which I think is the proper fix.
> > 
> > No, it is the *wrong* thing to do.  If you do stty -onlcr you should get
> > just one NL, nothing else.  It's the job of the tty layer and the serial
> > console write function to translate NL to CRNL, nobody else should mangle
> > the output in any way.
> 
> Right, that's why I'm saying that the l2 shouldn't mangle output, nor
> should the PROM.  They should just pass through whatever the serial
> driver hands them.  I think we're in agreement here?

Sorry, I should have been clearer:  our L2 system controller will be
fixed to *not* add CRs, as will our PROM, so we'll need your fix.
David, can you please apply it?  Sorry for the runaround.

Andreas, does that sound right?

Thanks,
Jesse

Re: Fix staircase effect on Altix serial console.

[David Mosberger]

>>>>> On Wed, 3 Mar 2004 14:21:16 -0800, jbarnes@sgi.com (Jesse Barnes) said:

  Jesse> Sorry, I should have been clearer:  our L2 system controller will be
  Jesse> fixed to *not* add CRs, as will our PROM, so we'll need your fix.
  Jesse> David, can you please apply it?  Sorry for the runaround.

If Andreas resends it, sure.

	--david

Re: Fix staircase effect on Altix serial console.

[Andreas Schwab]

jbarnes@sgi.com (Jesse Barnes) writes:

> Andreas, does that sound right?

Yes, thanks.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."