SUMMARY (almost): "Fast Data Access MMU Miss" when booting

From: NA/LMD/T/IA Roar Smith (
Date: Fri May 15 1998 - 11:17:56 CDT


here's a partial summary of responses and findings regarding my problem with
"Fast Data Access MMU Miss" when booting.

The responses:
============== (Brad Young ) had occasional erros of the same kind, but
no solution.

Janet Hoo <Janet.Hoo@Ebay.Sun.COM> wrote:

> Your motherboard is bad. : ( MMU is a memory management chip that sits on the
> board itself.

This was considered very unlikely by our local Sun reseller.

Bob Reitinger <> suggested the swap file was incorrect, but
the system never got as far as loading the kernel, so this was not the answer.

Our local Sun reseller suggested patch 103934, but 103934-06 was already
installed, and could not be the answer since the kernel is never loaded.

Further testing done by myself:

It turns out that booting is only possible from two out of four Differential
U-SCSI controllers, and has nothing to do with the D1000 disk arrays or with
the disks. When the cloned disk or the entire array was moved to one of the
two "proper" controllers, booting went just fine!

Both the bootable controllers are in the first SBus slot on the first SBus
on its I/O board (the middle slot named "1") - there are two I/O boards.

Both the un-bootable controllers are in the first SBus slot on the second SBus
on its I/O board (the left slot named "0").

This *could* be due to two defective controllers, but I think it much more
likely that the OpenBoot version only supports booting from the first SBus,
which is what I have suggested to our local Sun reseller - they have yet to
respond to this suggestion.

Roar Smith

The original posting:

------------- Begin Included Message -------------
>From lmdrsm Mon May 4 10:12:55 1998
Subject: "Fast Data Access MMU Miss" when booting
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-MD5: UtXTF/M+2Nd7gD5Rhl3Z/Q==


when trying to boot my Enterprise 4000 server from a clone of the
real boot partition, I get the following error from the boot prompt:

{0} boot /sbus@3,0/QLGC,isp@1,10000/sd@0,0
Boot device: /sbus@3,0/QLGC,isp@1,10000/sd@0,0 File and args:
Fast Data Access MMU Miss
{0} ok

This problem is not specific to that disk - I have tried the same
thing with another disk on another controller.

Background information:

The server is a Sun Ultra Enterprise 4002 with 4x336MHz/$MB CPU's,
two SBus I/O Boards each with two Ultra-SCSI controllers, on each of
these Ultra-SCSI controllers hangs a Sun Storedge D1000 with 12x 4GB
10.000 RPM drives (Seagate Cheetah drives) .

The operating system in Solaris 2.5.1 HW 11/97 with (I believe) all relevant
patches, DiskSuite 4.1 (SDS) and Sun Enterprise Volume Manager 2.5 (SEVM).

The /, /usr, /var, and /opt filesystems as well as swap are mirrored
with DiskSuite and the system will happily boot from any of the two
boot disks.

In order to be able to repair the system easily if it should crash
horribly at an unopportune time, I created a bootable drive on another
disk containing root, /usr, /var and /opt in one filesystem, and I could
actually boot from this drive several times, but after I made some changes
to the system, including but not limited to configuring SEVM, it is not
possible to boot from this drive anymore, nor have I been able to re-create
such a drive.

Here are the different things I have tried:

Approach A:
1. copy the existing root partition with "dd" (skip the first 8k)
2. fsck the copied partition
3. expand the new root filesystem with growfs
4. copy the /usr, /var and /opt filesystems with ufsdump | ufsrestore
5. edit etc/vfstab and etc/system with the metaroot command
6. install boot block with the installboot command

Approach B:
1. create a new root filesystem with newfs
2. copy the /, /usr, /var and /opt filesystems with ufsdump | ufsrestore
3. edit etc/vfstab and etc/system with the metaroot command
4. install boot block with the installboot command

Regardless of the approach I never get past the "Fast Data Access MMU Miss"
error, and I have even tried power-cycling the system and storage arrays
to no effect.

Anyone got any ideas?

Thanks in advance,

Roar Smith, M.Sc.E.E.
IT Architect L. M. Ericsson A/S
Phone: +45 3388 3577 Sluseholmen 8
FAX: +45 3388 3139 DK-1790 Copenhagen V

------------- End Included Message -------------

This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:12:40 CDT