Re: merge for pciide at cardbus?

[David Young <dyoung%pobox.com@localhost>]

On Thu, Oct 02, 2008 at 08:05:15PM -0700, Tom Spindler wrote:
> > I put pciide_cardbus that had done some corrections.
> > 
> > ftp://ftp.netbsd.org/pub/NetBSD/misc/kiyohara/tmp/pciide_cardbus_support-20080912.diff
> > ftp://ftp.netbsd.org/pub/NetBSD/misc/kiyohara/tmp/satalink_cardbus-20080912.diff
> > 
> > However, I have not tested this enough yet.  Can anyone test?  ;-)
> 
> I had to use patch -N, as there was some duplication between the two
> patchfiles. However - modulo the original 'I get an NMI upon probe'
> problem which seems to be somewhat related to my hardware - it works
> just fine for me.
> 
> (I'd like to know wtf the NMI thing occurs on one laptop but not
> on the other, however.)

The NMI occurs when your laptop's BIOS enables NMI for PCI exceptions on
the host bridge, and either there is a defect in your CardBus bridge or
in our driver for it, or the kernel tries to read or write a PCI register
but no device responds.

More often than other x86 BIOS, laptop's BIOS enables the NMI.  I do
not know why.

A backtrace can help us pin the cause of an NMI on a hardware/software
defect.  If you read any register, or if you write an I/O or configuration
register, the NMI ordinarily occurs at the very instruction that
produced the read or write.  However, if you write to PCI memory, the
write may be "posted," and then the NMI may come some arbitrary number
of instructions later.

Dave

-- 
David Young             OJC Technologies
dyoung%ojctech.com@localhost      Urbana, IL * (217) 278-3933 ext 24