This is still a problem for me.
I have had no support or help in any way on this one.
Hugh Bragg wrote:
>
> I'm running S/W raid-1 with 2 18G seagate barracuda on a Initio
> 1060P U2W SCSI Controller and a C&T 69000 VGA controller for
> AGP interface on Linux RH 6.2.
>
> Initially this works fine, untill I start using S/W raid-1.
> My XFont-server fails immediately.
> If I already have XFree86-3.3.6-20 running it starts cycling
> to get the display going but never does.
> If I attempt to run XF86Setup it freezes the computer tighter
> than a ...
>
> I have several of these boxes and the same thing occurs on all
> of them.
>
> The only possible explanation I can think of is that the SCSI
> controller and the VGA controller are on the same interrupt,
> but there is no way to change that as they are both embedded
> into the Peak 635 series motherboards, and I don't think this
> should make any difference.
>
> Does anyone know if the XFree86 or XFont-server somehow try to
> use the interrupt and possibly hang the SCSI controller?
>
> Is there any other explanation I can try?
>
> Hugh.
The manufacturer says the interrupt conflict isn't an issue
on an industrial machine so I wonder why Xwindows breaks when
raid is installed.
diana gray wrote:
>
> Hi Hugh,
>
> Yes in a normal (desktop) environment an IRQ is a system wide resource,
> however in an industrial environment where you may have up to 20 PCI slots
> and if you include onboard devices up to about 26 PCI devices, you require a
> revised architecture because you have long since exhausted your interrupts.
> The Processor is mounted onto a Single Board Computer (SBC), which has all
> the necessary controllers to form a complete system, in addition it also
> requires the ability to control a number of other PCI devices this is done
> by using PCI Bridge chips these can contain a number of bridges and
> essentially provide you with additional PCI-Bus's. Each BUS must contain
> devices with unique addresses but you may utilize the same address on
> multiple Bus's thus enabling you to address a much larger array of cards
> within your system.
> So esssentialy Passive Backplane architecture is very simular to the desktop
> enviroment but was specificaly designed to allow you to address more
> devices, a major draw back with desktop systems.
>
> I can only suggest that perhaps you could try to do a software RAID on the
> IDE channel. If it encounters the same problems, then it must be a software
> issue, which unfortunately I cannot support. I will however try and
> configure a similar set-up here and see if I can offer you any further
> assistance.
>
> RedHat Linux supports the Initio adapter, just not on the install disk.
>
> I wrote to Initio and they send me the install/boot images I required
> to get the device installed.
>
> I'm not too sure about the architecture of the bus systems on the
> Tigeray or how interrupts are directed onto the bus. I do know that
> an interrupt is a system wide resource and that unless the VGA adapter
> doesn't use one, it must have a unique interrupt. Some vga adapters
> support interrupt sharing but I doubt this is well supported widely as
> it must be controlled from the bios itself and it is difficult to
> predict how software will call these devices.
>
> It is difficult to understand why people go to all the trouble of making
> a bus programmable via a slot and then remove this programmability by
> embedding the device so that it has no slot. Surely the clever thing to
> do would be to allow the device to be programmable with the bios.
>
> We really want this interrupt to be moved away from the other device.
>
> The problem may not be with the interrupt conflict at all but there is
> no way of knowing this without trying.
>
> My only other alternative is to disable the Initio SCSI device and
> attach the disks to another scsi controller. That way I can be sure that
> the problem if it reoccurs has nothing to do with shared interrupts
> and something to do with the way Linux software raid interacts with
> the vga controller.
>
> Symptoms:
> Everything works fine until I install software raid-1. Upon doing so
> the Xfont server stops working and the X system cycles trying to
> initialise itself without succeeding. Attempting to run XF86Setup causes
> the system to hang and nothing short of a full reset brings it back.
> I have to run the system in run level 3.
>
> Help please.
>
> Hugh Bragg.
Does anyone have any idea how I might repair the problem as I'm unable
to
install oracle properly of use the normal configuration tools.
Hugh.
