Subject: Re: kernel for IQ80310 crashes...
To: John Clark <email@example.com>
From: Jason R Thorpe <firstname.lastname@example.org>
Date: 01/15/2002 20:16:07
On Tue, Jan 15, 2002 at 08:02:25PM -0800, John Clark wrote:
> > Hm, did you load this via the Ethernet, or copy it into flash and boot
> > it from there?
> System setup:
> Board: IQ80310
> Backplane: (Cheating) ATX mother board with no CPU installed.
> PCI Slots: 2 used: (arbitrary numbering):
> Slot 1: LSI Logic PCI card with 53c1010 SCSI controler.
> Slot 2: PCI card with Intel 82544GC Gigabit Ethernet.
> (I'm not trying to get either of these PCI devices up at the moment,
> there should be no drivers in the kernel for them.)
Ah, okay, this is an actual IQ80310 board, then, rather than a board
based on the IOP310 design?
Note, it's unlikely that devices in the IQ80310 slots will work in any
meaningful way, due to a fundamental design flaw in the IQ80310's interrupt
routing logic. Basically, the problem is that interrupts from those PCI
slots can't be blocked without blocking *all* interrupts. If you really
need to be able to use those slots, then I'll see if there's something I
can do to hack around the problem.
> I downloaded the image netbsd-fxp0 built straight from the cvs archives,
> burnt that into flash, then before the above
> 'go' command, had copied it back out of flash into ram.
Hm. This is something I haven't tried yet -- I have just been loading
the kernel from the network each time... I'll make sure this also works
> These are eval boards. But I think the doc says that copper gig-e device
> is 0x100d.
> Do you have a fiber board?
No, all my Cordova boards are copper (I have some i82544EI and i82544GC
boards -- the EI boards are PRO/1000 XT Server Adapters, and GC boards
> The memory starts at 0xa0000000, and I have 128 Megs rather than the 32
> which is the 'stock' board.
Hm. I'll get a 128M stick for mine and see if mine fails in the same way.
-- Jason R. Thorpe <email@example.com>