Re: MVME2604 Problems with MMU

Gabriel Paubert [off-list ref] writes:

On Thu, 9 Nov 2000, fraher wrote:

quoted

   I've got a 2.2.12 kernel with Gabriel's patches (NFS mounted root),
and it works beautifully on a MVME2603 board (many thanks, Gabriel).
However, when I swap out the MVME2603 with a MVME2604, and cause some
SCSI disk activity, the system crashes with

        end_scsi_request: buffer-list destroyed
        page fault in interrupt handler, addr=8000000
        NIP: 08000000 XER: 20000000 LR: C00C67BC REGS: c00ffc48 TRAP: 0400
        MSR: 40001032 EE: 0 PR: 0 FP: 0 ME: 1 IR/DR: 11
        TASK = c00fdfe8[0] 'swapper' mm->pgd c00fc000 Last syscall: 112
                ...
        kernel panic: kernel access of bad area pc 8000000 lr c00c67bc address 8000000 tsk swapper/0

The 'page fault' makes me wonder if the 604's TLB miss response differs
sufficiently to cause the kernel panic.
Has anyone else seen problems running on 2.2.12 with an MVME2604?
Are there other patches needed to compensate for the MMU differences
with the 604e?

I don't have any MVME2604, but I have several MVME2400 with uptimes in
the months range. MVME2400 don't have SCSI, but a 750 has processor with
an MMU very similar to the 604e. Could it be a subtle SCSI/timing problem ?

OTOH, I have reports of working 2604, but I don't remember if it's with
root on NFS or using SCSI.

	Gabriel.

I have seen similar errors on both 2604 and 2400, but the errors on
the 2604 were quite some time ago.

The 2400 problems happened during heavy IDE activity and have
disappeared when I upgraded to 2.2.17+debian patches.

I thought they were related to my IDE problems and didn't look in to
the further.

I will probably try 2.2.17 on some 2604 cards this weekend. I'll
report if I see similar problems.

/Micke

--
Michael Lundkvist
ml@epact.se

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/