ttyACM0: when the device dies, 2.6.30-rc7 dies, too (regression?)

ttyACM0: when the device dies, 2.6.30-rc7 dies, too (regression?)

[Pavel Machek]

Hi!

I believe this worked before; on 2.6.30-rc7, when ttyACM0 dies
(openmoko in bootloader powers off) the notebook I'm running minicom
on dies, too :-(. For the first time, I did not catch any messages,
I'll retry the test on console.

									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: ttyACM0: when the device dies, 2.6.30-rc7 dies, too, strange echo (regression)

[Pavel Machek]

Hi!

> I believe this worked before; on 2.6.30-rc7, when ttyACM0 dies
> (openmoko in bootloader powers off) the notebook I'm running minicom
> on dies, too :-(. For the first time, I did not catch any messages,
> I'll retry the test on console.

Nothing useful on the console...

And it gets worse: with 2.6.30-rc7, I only get echo on character N
when I type character N=1. Which is extremely confusing and basically
makes it unusable for interactive work. This definitely worked
before... but it was quite long ago.

									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: ttyACM0: when the device dies, 2.6.30-rc7 dies, too, strange echo (regression)

[Oliver Neukum]

Am Montag, 1. Juni 2009 12:25:46 schrieb Pavel Machek:
> Hi!
>
> > I believe this worked before; on 2.6.30-rc7, when ttyACM0 dies
> > (openmoko in bootloader powers off) the notebook I'm running minicom
> > on dies, too :-(. For the first time, I did not catch any messages,
> > I'll retry the test on console.
>
> Nothing useful on the console...
>
> And it gets worse: with 2.6.30-rc7, I only get echo on character N
> when I type character N=1. Which is extremely confusing and basically
> makes it unusable for interactive work. This definitely worked
> before... but it was quite long ago.

Do you have any reason to assume these bugs are related? If not,
please keep them separate.

Can you please be more specific what exactly fails?

	Regards
		Oliver

Re: ttyACM0: 2.6.30-rc7 dies on device disconnect, strange echo (regressions)

[Pavel Machek]

Hi!

First, thanks for quick reply.

> > > I believe this worked before; on 2.6.30-rc7, when ttyACM0 dies
> > > (openmoko in bootloader powers off) the notebook I'm running minicom
> > > on dies, too :-(. For the first time, I did not catch any messages,
> > > I'll retry the test on console.
> >
> > Nothing useful on the console...
> >
> > And it gets worse: with 2.6.30-rc7, I only get echo on character N
> > when I type character N=1. Which is extremely confusing and basically
> > makes it unusable for interactive work. This definitely worked
> > before... but it was quite long ago.

> Do you have any reason to assume these bugs are related? If not,
> please keep them separate.

Well, ok, they are probably two different bugs...

> Can you please be more specific what exactly fails?

...but it is quite easy to test for both of them and they seem to be
introduced in similar timeframe.

Steps to reproduce:

minicom /dev/ttyACM0

Bug #2: try to interactivelly communicate with the other
device. Notice that characters are echoed one character too late. You
type "hello", but only "hell" appears. You press "!", "o" appears. 

Bug #1: while minicom is still connected, pull the USB cable. Machine
locks hard.

Now, both bugs are present in 2.6.28, and 2.6.24-rc8 does not have
either problem.

It would be nice if someone could verify if it also happens with
regular USB modem -- I'm using openmoko to reproduce it.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: ttyACM0: 2.6.30-rc7 dies on device disconnect, strange echo (regressions)

[Oliver Neukum]

Am Montag, 1. Juni 2009 13:31:51 schrieb Pavel Machek:
> Steps to reproduce:
>
> minicom /dev/ttyACM0
>
> Bug #2: try to interactivelly communicate with the other
> device. Notice that characters are echoed one character too late. You
> type "hello", but only "hell" appears. You press "!", "o" appears.
>
> Bug #1: while minicom is still connected, pull the USB cable. Machine
> locks hard.

I can replicate neither bug with 2.6.30-rc7. Can you replicate it with
anything but openMOKO?

	Regards
		Oliver

Re: ttyACM0: 2.6.30-rc7 dies on device disconnect, strange echo (regressions)

[Pavel Machek]

On Mon 2009-06-01 14:34:56, Oliver Neukum wrote:
> Am Montag, 1. Juni 2009 13:31:51 schrieb Pavel Machek:
> > Steps to reproduce:
> >
> > minicom /dev/ttyACM0
> >
> > Bug #2: try to interactivelly communicate with the other
> > device. Notice that characters are echoed one character too late. You
> > type "hello", but only "hell" appears. You press "!", "o" appears.
> >
> > Bug #1: while minicom is still connected, pull the USB cable. Machine
> > locks hard.
> 
> I can replicate neither bug with 2.6.30-rc7. Can you replicate it with
> anything but openMOKO?

So AT commands echo properly on USB modem or something? Bad :-(.

I'm afraid I do not have another ACM device near me. I guess going
through cdc-acm.c 's history would be relevant?
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: ttyACM0: 2.6.30-rc7 dies on device disconnect, strange echo (regressions)

[Oliver Neukum]

Am Dienstag, 2. Juni 2009 11:19:13 schrieb Pavel Machek:
> On Mon 2009-06-01 14:34:56, Oliver Neukum wrote:
> > Am Montag, 1. Juni 2009 13:31:51 schrieb Pavel Machek:
> > > Steps to reproduce:
> > >
> > > minicom /dev/ttyACM0
> > >
> > > Bug #2: try to interactivelly communicate with the other
> > > device. Notice that characters are echoed one character too late. You
> > > type "hello", but only "hell" appears. You press "!", "o" appears.
> > >
> > > Bug #1: while minicom is still connected, pull the USB cable. Machine
> > > locks hard.
> >
> > I can replicate neither bug with 2.6.30-rc7. Can you replicate it with
> > anything but openMOKO?
>
> So AT commands echo properly on USB modem or something? Bad :-(.

My cell phone echoes back arbitrary strings.

> I'm afraid I do not have another ACM device near me. I guess going
> through cdc-acm.c 's history would be relevant?

A bisection would indeed be very good.

	Regards
		Oliver

Re: ttyACM0: 2.6.30-rc7 dies on device disconnect, strange echo (regressions)

[Oliver Neukum]

Am Dienstag, 2. Juni 2009 11:19:13 schrieb Pavel Machek:
> > I can replicate neither bug with 2.6.30-rc7. Can you replicate it with
> > anything but openMOKO?
>
> So AT commands echo properly on USB modem or something? Bad :-(.
>
> I'm afraid I do not have another ACM device near me. I guess going
> through cdc-acm.c 's history would be relevant?

The tty layer was changed a lot, so going only through acm's history
is risky. I suggest you do a full bisection.

	Regards
		Oliver