Re: Disabling ACPI also implicitly disables SMP support for Intel Atom chips?

["Steven M. Bellovin" <smb%cs.columbia.edu@localhost>]

On Wed, 04 Mar 2009 23:27:55 -0500
"David H. Gutteridge" <dhgutteridge%sympatico.ca@localhost> wrote:

> On Wed, 04 Mar 2009 22:25:35 -0500, David H. Gutteridge wrote:
> >The main problem I've found so far with ACPI being enabled (I've only
> >been using this machine a few days) is that the Ethernet card doesn't
> >work.  With ACPI, the card doesn't seem to transmit any packets, and
> >every few seconds I see the following output to the console:
> >
> >re0: watchdog timeout
> >re0: reset never completed!
> >
> >If I disable ACPI, the re(4) card works fine.
> 
> Following up on my own comments, I found another NetBSD user's
> comments instructive[1].  Martijn van Buul had written: "I'm having
> problems with either my network adapter, or my keyboard. I get to
> choose; if I disable ACPI, my ethernet adapter will work just fine
> but my keyboard won't, if I enable ACPI my keyboard will work, but my
> ethernet won't. The peculiar thing is that *both* worked at the same
> time during installation."  (In his case he's using an Acer Aspire
> One.)  Like me, his dmesg has the "watchdog timeout" errors at the
> end, and like me, he's using an re(4) card.
> 
> I thought back to my install, and remembered I'd noted the same thing:
> the GENERIC kernel I'd booted during install had ACPI enabled, I
> believe.  The re(4) card had worked fine then for me too, as I'd
> transferred files via ftp with it.  I guessed that somehow configuring
> the card during the boot process rather than afterward was triggering
> this "watchdog timeout" problem.
> 
> So I removed my /etc/ifconfig.re0 file, rebooted with ACPI on,
> manually brought up the interface, manually added the default route,
> and voila, re(4) works fine.  So that seems to be the workaround for
> people who absolutely want both HT and re(4) on these little devils
> (pun intended when they're running a BSD).
> 
Hmm...  You implied you were using X, as I recall.  One thing I've seen
when starting up X is that interrupts are disabled for too long during
startup, causing "fatal firmware errors" on wpi0.  Until I started
running wpa_supplicant, I had to do 'ifconfig wpi0 down up' once I was
in X.  I wonder if you're seeing something similar.


                --Steve Bellovin, http://www.cs.columbia.edu/~smb