2.2.20 oops during fsck, mylex 352 (DAC960), CONFIG_3GB

[Hank Leininger]

Are there any known issues with 2.2's deliberately-partially-disabled 3GB
physical-RAM support (#define PAGE_OFFSET_RAW 0x40000000 in page_offset.h)
and either the DAC960 driver, or e2fsck v1.22 ?

I've ended up (because of a comedy of RMA-related errors) with 3 1GB DIMMs
for a new server, when I'd originally planned only 2.  For several reasons
I'd like to stick to 2.2 for a bit longer.  But to get 3GB support in 2.2 I
had to "complete" the CONFIG_3GB config option, and thinking that I knew
what I was doing, started smacking the box around.  Everything seemed
solid, until I killed the box with 8k NFS (I think--was getting
>10mbytes/sec writes before the lockup though :).  That's not the problem
I'm worrying about now, though :-P

Upon reboot+fsck, while I cursed myself for not yet having converted to
ext3, the box oops'ed.  It isn't local to me and doesn't have a serial
console yet, so it was rebooted w/o recording the oops info.

Now, I'm *hoping* there's a known or easily-identifiable issue with either
the DAC960 driver, or (less likely) with e2fsck, when running with
PAGE_OFFSET_RAW 0x40000000.  If so, then I'll take the doctor's advice and
won't do that.  If not, then I may have *another* problem[1].  So, someone
please tell me this is a known problem ;)  OTOH, glancing through source
for the DAC960 driver and for fsck I don't see anything that looks like it
would care about PAGE_OFFSET.  And, CONFIG_2GB seems to be solid.  Is
anyone running a similar configuration who has *not* had these problems?

[1] I will first suspect a hardware problem.  This is all for a replacement
    server to upgrade the one that runs MARC.  In the 4 months since I
    bought all the hardware, nearly half of the components--all high-end,
    supposedly high-quality, certianly expensive--have failed, or were
    simply DOA.  I hate hardware.  

Thanks,

Hank Leininger <[e-mail]>