Subject: Re: amd64 reboot when using acpi
To: Brett Lymn <>
From: Chavdar Ivanov <>
List: current-users
Date: 06/20/2007 13:19:40
On 20/06/07, Brett Lymn <> wrote:
> On Wed, Jun 20, 2007 at 12:24:54PM +0100, Chavdar Ivanov wrote:
> >
> > Reserving IRQ7 in the BIOS did not help.
> >
> Interesting.  It seems to have worked for me (if I remember) try
> applying the attached patch and see if that makes things better - it
> tries to nuke pending interrupts when the ehci driver takes control of
> the device.
> > In the mean time I got another panic - this time in OHCI while copying
> > large ~ 4GB - directory from an external USB disk, after copying some
> > 500MB:
> >
> That's odd - I have not done any bulk transfers over usb yet.
> Your issue with the nfe is strange too - both my nfe interfaces work
> fine, quite a few times I have transferred ~300Mb files between my
> laptop and my main machine connected with a crossover cable.  I have
> never seen wire speed but the transfer rate is quite respectable - in
> the region of 500 - 600 Mbit/s

The interfaces work fine when ehci is disabled; they stop if I disable acpi.

> One thought, what is the revision of your bios?  I am running v1.04 at
> the moment.  I found that v1.05 caused a non-acpi kernel to hang and
> the acpi kernel would still crash at the same place.  Now that I can
> boot an acpi kernel I may just update the bios to the latest.

I am using 1.05 as at one stage I had two GF 8800GTX cards in SLI,
which was not possible with 1.04. I might downgrade to 1.04 to see.

There may be some other problem on this m/b - NetBSD apparently
correctly uses both nfe interfaces, but I have never been able to get
the bottom one to work on any version of Windows - from XP/32 SP2 to
LongHorn Beta3 64 bit - it always reports insufficient resources
available, free some...  The hardware was lying in state in our stock
room and no one knew if there was any problem with it or why it hasn't
been used in a customer machine, so I scrapped some bits from here and
there and built the box to have at least for testing around, only the
p/s and the case were new.

I am not too bothered though - I can live with no ehci here. It was
more interesting to find out if there is some underlying problem with
NetBSD amd64. What is a bit strange is that the panics happening with
ehci on sometimes lead to resets, sometimes to freezes, so it is not
quite consistent.

> --
> Brett Lymn

Chavdar Ivanov