Howto debug Kernel problems ? (jail -r -> freeze)

[Marius Strobl]

On Thu, Jan 26, 2012 at 12:55:02PM +0100, Denny Schierz wrote:
> hi,
> 
> Am 25.01.2012 um 16:53 schrieb Michiel Boland:
> 
> > sysctl debug.kdb.alt_break_to_debugger=1
> 
> great :-), it works (until ... )
> 
> I did the test again:
> 
> /etc/rc.d/jail onestart
> /etc/rc.d/jail/onestop
> 
> jls ...
> 
> jail -r 1
> 
> The command hangs, the server doesn't accept  ssh connections anymore, but ping works. But only until I make something like "telnet localhost 22" to check, if SSH accept from localhost ..  .... FREEZE ....
> 
> back to ALOM, send break ... connect ... nothing. -> Reset
> 

>From your email it's once again not clear whether the alternate
break sequence works or not when the the machine hangs. If it
doesn't the only thing I currently can think of is to resort to
printf(9)-debugging starting at sys_jail_remove() in kern_jail.c
in order to determine where the hang occurs. This somewhat smells
like the kernel trying to directly use and userland address and
not going through copy{in,out}(9), which would match both the
symptoms of a hang and it just working anyway on x86. On a
quick glance at kern_jail.c I can't spot an obvious issue. From
your previous emails and the mailing list reference you gave it's
also not clear whether you are using some kernel patch which might
introduce such a bug though.

Marius