On Tue, 2003-08-05 at 22:08, Andi Kleen wrote:
> On Tue, Aug 05, 2003 at 12:38:11PM -0700, Andrew Morton wrote:
> > Andi Kleen <ak@muc.de> wrote:
> > >
> > > Sometimes drivers do long things with interrupt off and the NMI watchdog
> > > triggers quickly. This mostly happens in error handling. It would be
> > > better to fix the drivers to sleep in this case, but it's not always
> > > possible or too much work.
> >
> > yup.
> >
> > Do we need an mdelay_while_touching_nmi_watchdog() variant?
>
> Maybe that would be too encoraging for broken code.
>
> Admittedly I did that by hand for the MPT fusion driver (which currently
> triggers the watchdog when it gets into any error handling situation)
> This especially hurts on x86-64 which runs the watchdog by default.
> But it's strictly a bad hack, not a good interface.
having a more generic/portable "trigger_watchdog" function would be
better then, such that ALL watchdogs, and not just the NMI one can hook
into this
This archive was generated by hypermail 2b29 : Thu Aug 07 2003 - 22:00:30 EST