ACPI is the strongest probable cause at the moemtn
ACPI: Power Button (FF) [PWRF]
ACPI: Power Button (CM) [PWRB]
ACPI-0339: *** Error: Looking up [\_PR_.CPU0] in namespace, AE_NOT_FOUND
search_node effc5ca0 start_node effc5ca0 return_node 00000000
ACPI-0339: *** Error: Looking up [FAN_] in namespace, AE_NOT_FOUND
search_node effc5ce0 start_node effc5ce0 return_node 00000000
They're cheap $75 non-raid cards, but have worked perfectly in another
machine.
I also get this in dmesg
-----Original Message-----
From: Volker Kuhlmann [mailto:[EMAIL PROTECTED]
Sent: Tuesday, 21 February 2006 10:35 a.m.
To: [email protected]
Subject: Re: DMA strangeness
> The performance is terrible. But when I enable DMA on the drives with
> hdparm -d1 /dev/hd[e-l]
> About 10 seconds later I get kernel messages like
> hde: lost interrupt
> And the only way to clear this is reboot.
Possibilities I can see:
Buggy kernel and/or buggy chipset with insufficient kernel workaround. This
is the IDE part of the mobo chipsets, and all IDE cards.
Disks interfering with each other if you have more than one on a cable.
(not likely) Bad cable connection.
ACPI / interrupt problems in your box. What are the IDE interrupts shared
with? Can you change that for a test?
> The hardware is a dual PII 400 with two promise PCI IDE controllers.
Ouch, stay clear of that Promise cr*p. AFAIAA, Promise does not have a
single product worth using. Use non-raid PCI IDE cards instead, or get
something decent (warning: 4 figure price). All that cheap raid shite is
software raid anyway and is for Microsofties without software raid in their
OS.
If you can't avoid such a raid card, use it in IDE mode only and ignore its
raid functionality.
> Everytime I restart it the raid rebuild begins from scratch too.
Interesting. What does that indicate? A disk having been taken offline? Just
a disk I/O error shouldn't cause this.
Volker
--
Volker Kuhlmann is possibly list0570 with the domain in
header
http://volker.dnsalias.net/ Please do not CC list postings to
me.
