here is a nice oops copied down by hand.
It seems to be due the physical != kernel memory change, or better say
some physical = kernel mem assumption in ncrBsd.
ncr53c810-0: using memory mapped IO at virtual address 0xc2821f00
Unable to handle kernel paging request at virtual address 82821f04
current->tss.cr3 = 00101000, %cr3 = 00101000
*pde = 00000000
Oops: 0000
CPU: 0
EIP: 0010:[<c01a0cbd>]
EFLAGS: 00010202
eax: c2821f00 ebx: 00000f00 ecx: 00316cb0 edx: 60000000
esi: c0314068 edi: ffbe7f00 ebp: c0314018 esp: c031eebc
ds: 0018 es: 0018 fs: 002b gs: 0018 ss: 0018
Process swapper (pid: 1, process nr: 1, stackpage = c031e000)
Stack: c01ce9c8 00000001 00000001 ffffffff c2821000 00000f80 c00f7001 00000001
c01a5558 c01cca5c 00000000 00000001 00000001 0000032a ffbe7f00 0000fc00
0000000c 00000000 00000070 00000000 00000000 00000000 00000000 c01c000f
Call trace: [<c2821000>] [<c01a5558>] [<c01a522d>] [<c0199fe0>] [<c0198344>] [<c016f379>] [<c012d830>]
[<c010a535>] [<c011e184>] [<c0109386>] [<c01091b4>] [<c01091bb>] [<c010933c>]
Code: 8a 80 04 00 00 c0 eb 0f 8d 76 00 66 86 96 78 0f 00 00 66 83
Using `System.map' to map addresses to symbols.
>>EIP: c01a0cbd <ncr_attach+209/524>
Code: c01a0cbd <ncr_attach+209/524> movb 0xc0000004(%eax),%al
Code: c01a0cc3 <ncr_attach+20f/524> jmp c01a0cd4 <ncr_attach+220/524>
Code: c01a0cc5 <ncr_attach+211/524> leal 0x0(%esi),%esi
Code: c01a0cc8 <ncr_attach+214/524> xchgb %dl,0xf78(%esi)
Code: c01a0ccf <ncr_attach+21b/524> addw $0xffffff90,(%eax)
Code: c01a0cd3 <ncr_attach+21f/524> nop
Code: c01a0cd4 <ncr_attach+220/524> nop
ncr_attach source contain loads of mmaping stuff, i let Gerard look at it.
PS: strange that it works on alpha.
Context: P90, bargain ncr810 adapter, 32MB edoram, triton I, gcc2.7.2.l.3,
pre 2.1 applied to plain 2.0.21
I keep the ncr53c8xx.o, the kernel et al in case you need further details.
Nice to have a *real* oops, it's been so long :*>
-- Philippe Strauss, CH-1092 Belmont Email: philippe.strauss@urbanet.ch