Re: PMU, was Re: d-i built (but not tested)
On Sat, 13 Aug 2011, Wouter Verhelst wrote:
> On Fri, Aug 12, 2011 at 09:15:39PM -0700, Brad Boyer wrote:
> > On Sat, Aug 13, 2011 at 11:37:58AM +1000, Finn Thain wrote:
> > > On Fri, 12 Aug 2011, Brad Boyer wrote:
> > > > The basic idea is that any time the PMU thinks you aren't following the
> > > > right communication protocol (including the timing between messages) it
> > > > will just shut off because it thinks the OS crashed. So if it is in a
> > > > multi-message handshake and something disables interrupts for too long,
> > > > the system will just power off.
> > >
> > > So the solution would be to load the PMU driver earlier?
> >
> > That might be the issue. I'm surprised it's not loaded earlier than
> > the step where it failed.
>
> If it's not compiled in
It is present in your kernel binary (and indeed can't be built as a
module) but it is buggy.
The PMU driver is started by adb_init() which is a device_initcall, but
this happens too late. I worked around the problem successfully by making
it an arch_initcall instead.
Finn
> it would be in the initramfs. If it's in the initramfs, we need to
> unpack it before it can be loaded...
>
>
Reply to: