Curious BeagleBone console hang...

[Tim Kientzle]

On Apr 20, 2014, at 5:35 PM, Ian Lepore <ian at FreeBSD.org> wrote:

> On Sun, 2014-04-20 at 16:37 -0700, Tim Kientzle wrote:
>> Here's an odd one:
>> 
>> $ stty 72
>> 
>> consistently hangs the serial console on a BeagleBone
>> running FreeBSD-CURRENT r264208.  (I found this out
>> by accidentally omitting 'rows' when I typed the command.)
>> 
>> Doesn't seem to affect an SSH session, only the serial console.
>> 
>> After doing this, nothing seems to be running; the console
>> is just completely non responsive: even echo is dead.
>> 
>> I can recover by logging in through SSH and killing the
>> console login.
>> 
>> Trying to puzzle it out:
>> 
>> $ truss stty 72
>> ...
>> ioctl(0,TIOCGETA,0xbffffcb0)                     = 0 (0x0)
>> ioctl(0,TIOCGETD,0xbffffc9c)                     = 0 (0x0)
>> ioctl(0,TIOCGWINSZ,0xbffffcdc)                   = 0 (0x0)
>> ioctl(1,TIOCGETA,0xbffffb60)                     = 0 (0x0)
>> ioctl(2,TIOCGETA,0xbffffb60)                     = 0 (0x0)
>> fstat(1,{ mode=crw------- ,inode=32,size=0,blksize=4096 }) = 0 (0x0)
>> fstat(2,{ mode=crw------- ,inode=32,size=0,blksize=4096 }) = 0 (0x0
>> 
>> 
>> Does this make sense to anyone?
>> 
>> Tim
> 
> You're setting the line speed to 72 baud, and due to recent kernel
> changes the console baud rate is allowed to be changed now, so you lose
> console comms as soon as the change takes effect.

Thank you!

I went through the stty man page several times
and consistently overlooked the description for
a bare number.

That makes perfect sense.

I guess getty must reset the line when the
login session exits, which would explain why
killing the login session recovers it.

Tim