Mailinglist Archive: opensuse (2348 mails)

[Anders Johansson <ajh@xxxxxxxxxx>]

On Friday 04 April 2008 18:40:55 Randall R Schulz wrote:
> On Friday 04 April 2008 09:39, Anders Johansson wrote:
> > On Friday 04 April 2008 18:32:57 Randall R Schulz wrote:
> > > On Friday 04 April 2008 09:27, David C. Rankin wrote:
> > > > ...
> > > >
> > > > > Anders
> > > >
> > > >         Well, your right as usual about my understanding. What I can't
> > > > get my mind around is the fact that the mce(s) go away as long as
> > > > the driver isn't loaded. ...
> > >
> > > The driver tells the hardware, among other things, where to
> > > transfer data via DMA. Such transfers are done asynchronously and
> > > if the driver gives bad addresses or transfer sizes to such a
> > > DMA-equipped card, then that card's actions (carrying out the
> > > driver's instructions) can cause "hardware" errors.
> >
> > I don't think that will trigger an MCE though
>
> Well, one thing that can happen is that interrupt vectors can be
> clobbered by DMA activity, after which it can become impossible for the
> processor to continue because it gets a fault trying to execute from
> bad addresses or from valid addresses that contain gibberish
> instructions.

Well, it's obviously possible for software to hang a CPU, that's not in 
question.

But I still don't think it will trigger a machine check exception. That's not 
the type of errors those are designed to alert about

Anders
-- 
Madness takes its toll
-- 
To unsubscribe, e-mail: opensuse+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: opensuse+help@xxxxxxxxxxxx