Re: serial ttys broken?

On Tue, Sep 20, 2005 at 06:48:39PM +0100, Gavin Atkinson wrote:
> FreeBSD/i386 (buffy.york.ac.uk) (ttyd0)
> 
> login:
> FreeBSD/i386 (buffy.york.ac.uk) (ttyd0)
> 
> login: HWV*Q)kB
>                BJjHVKCCeeBSD/i386 (buffy.york.ac.uk) (ttyd0)
> 
> login:
> FreeBSD/i386 (buffy.york.ac.uk) (ttyd0)
> 
> login: CCreeBSD/i386 (buffy.york.ac.uk) (ttyd0)
> 
> login:
> FreeBSD/i386 (buffy.york.ac.uk) (ttyd0)
> 
> login: CCeeBSD/i386 (buffy.york.ac.uk) (ttyd0)
> 
> login:
> 
> 
> Yes.
> 
> And that's with ttyd0, not cuaa0, and a Windows machine at the far end
> of a 9600 baud link.  A configuration which I'm pretty sure should
> perform better.  No matter how hard I try, I can't manage to reproduce
> the corruption once logged in, unless I explicitly
> run /usr/libexec/getty from the shell, in which case the corruption is
> even worse:
> 
> loi:cCreeBSDi36(ff.or.c.(tt0
> 
> loi:
> reeBSDi36(ff.or.c.(tt0
> 
> loi:
> reeBSDi36(ff.or.c.(tt0
> 
> loi:
> reeBSDi36(ff.or.c.(tt0loi:
> reeBSDi36(ff.or.c.(tt0
> 
> loi:cCeeBSDi36(ff.or.c.(tt0
> 
> 
> Guess there may be some issue with getty?

Very interesting, I saw the same with multiple machines (alpha and i386).
My first thought was having multiple getty running on the same tty,
but this turned out to be wrong.
My second thought was that there is a bug with my ubser(4) serials, but
since it wasn't reproduceable with suns and after login it sounded
unlikely.
In some cases a single return could leed to garbadge, one or more cases
even endless flooded.
Now I know for shure that this is a FreeBSD problem.

-- 
B.Walter                   BWCT                http://www.bwct.de
bernd_at_bwct.de                                  info_at_bwct.de