[Ksummit-2012-discuss] [ATTEND] kernel core dump and "dying breath"

[Jiri Kosina]

On Wed, 27 Jun 2012, H. Peter Anvin wrote:

> > There is another thing that crossed my mind.  Think what we do with the
> > keyboard when we have one.  Totally predictable.  Hit page up, stare at
> > it for 30s, hit page up again, etc.  Meaning, that if we instead skip
> > the whole needing the input, and simulate it at panic, it could be a
> > nice win for minimal cost.  Page up every 30s.  Hit the top?  Page
> > down every 30s.  Repeat until power-off.  You could collect the whole
> > panic message with a camera too, albeit in chunks.
> 
> Mmmm... timers.

Well, I believe this could be worked around somehow. Very long cpu_relax() 
loops emulating the delays, etc (yes, the actual length of those would 
probably have to be pre-calibrated during early bootup phase, but that's 
easy).

-- 
Jiri Kosina
SUSE Labs