Subject: Re: port-i386/7807: 1.4 install floppy has "Bus-Master DMA error: missing interrupt" errs on install
To: Manuel Bouyer <>
From: Chris G. Demetriou <>
List: netbsd-bugs
Date: 06/23/1999 16:26:30
Manuel Bouyer <> writes:
>On Fri, Jun 18, 1999 at 02:29:46AM -0700, wrote:
>> pciide0:0:0: lost interrupt 
>>         type: ata
>>         c_bcount: 65536
>>         c_skip: 0
>> pciide0:0:0: Bus-Master DMA error: missing interrupt, status=0x21
>> wd0e: DMA error writing fsbn 129184 of 129184-129311 (wd0 bn 764224; cn 758 tn 2
>>  sn 34), retrying
>> wd0: soft error (corrected)
>> It's worth noting that these errors do appear to be safely corrected,
>> i.e. installation continues and nothing _appears_ to be wrong
>> with the installed bits (though i've not tried to verify them
>> closely enough to be sure).
>I'm pretty sure these errors are caused by flacky hardware which don't support
>DMA properly.

While this may be true, i'm hesitant to believe it.

The system in question is an effectively-new Digital box (effectively
because it was brand new but out of date when i bought it > a year
ago, and it's spent most of its time turned off), with the stock
from-factory hardware configuration except for the addition of the PCI
ethernet card.

>A too long cable, or with the wrong
>impedence can be the cause of this. Can you try to use another

I can't easily since I only own three and the other two are in use
(and one's the identical twin to this one.)

However, given that the machine is effectively-new and the cable is
the one shipped from the factory (i've never even unplugged it!), and
they presumably do QA on their hardware, i'd expect it to perform
properly.  Anyway, if I get a chance, i'll try to get another IDE cable.

>Also, if you have a 3-connector cable with only one disk, you must put the
>controller on one end and the disk on the other (i.e. with the empty connector
>in the midlle). 

It's a 3-connector cable, and the controller and disk are on the
ends.  This is how it was shipped from the factory.

> > Different block numbers are reported for the different errors within
> > a single install, and are reported for subsequent install attempts.
> Yes, random failures ... typically a hardware problem on the IDE line.

Does it mean anything that c_bcount always appears to be 64k and
c_skip always appears to be 0?

Chris Demetriou - -
Disclaimer: Not speaking for NetBSD, just expressing my own opinion.