Linus, this looks disturbingly familiar.
On Fri, 15 Jan 1999 11:47:51 +0100, Kurt Huwig <kurt@huwig.de> said:
> Am Fri, 15 Jan 1999 schrieben Sie:
>> On Fri, 15 Jan 1999, Kurt Huwig wrote:
>>> I had the attached Oopses while compiling.
>>
>> Interesting. "bh->b_next" is crap (it has the value 0xffff, which is not a
>> kernel pointer).
> I was the one with the nonstandard APM-BIOS. Ok:
I spent some time debugging a box with Doug Ledford and Mike Wangsmo.
They were able to reproduce buffer_head corruptions (specifically,
bh->b_next pointing to crap) during redhat installs, although the
machine worked perfectly once installed. We eventually narrowed it down
to what looked like a BIOS32 problem (although not APM in this case): it
appears that the BIOS32 was playing with page tables and failing to
invalidate the TLB on return. We did manage to come up with a modified
debugging kernel in which a single tlb_flush_all() after PCI setup was
sufficient, and necessary, to eliminate the symptoms.
There are some obvious workarounds for this sort of problem, but they
are all gross. (Incidentally, booting "mem=16m" also proved to be a
sufficient workaround in our particular case.)
Just another data point you may find useful. I really don't think the
problem was in Linux in the case we investigated.
--Stephen
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/