Reboot instead of debugger/dump from X11? (Re: an accidental way to pull the plug)

[Poul-Henning Kamp]

In message <Pine.NEB.3.96L.1041225213846.45272E-100000 at fledge.watson.org>, Robe
rt Watson writes:
>
>On Sat, 25 Dec 2004, Chuck Robey wrote:
>
>> I got a little too fast just now, and did something that I shouldn't
>> have done, but what's interesting is, FreeBSD decided to pull the plug
>> and instantly reboot, which is usually something to be regarded as
>> pathological, so I am letting you folks know.  I haven't analyzed
>> exactly what happened to cause it, actually. 
>
>With the recent 802.11 panic I posted about a bit earlier, I saw something
>rather odd: when the panic took place on the normal text console, the
>system dropped to the debugger.  However, when it took place with X11
>running, I got an "instant reboot" -- no dump, etc.  So something odd is
>happening during even normal failure modes when X11 is running for me.  I
>wonder if anyone else is seeing this?

Well, it's worse:  I often get two-level panics these days and I have
no idea how to find out what the first panic was, panicstr isn't even
set until after some SMP magic which appearantly is responsible for
the second layer of panicing.

I think we should start panic() with setting panicstr with an atomic
instruction so that we always get the first panic recorded, no matter
what.

I suspect the preemption code is at fault.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.