Re: Potato woes
Hello again.
On Fri, Apr 23, 1999 at 03:18:01PM -0500, Michael Stroucken wrote:
> > Moving to a 2.2.6 kernel seems to fix the date problems but due to
> > multiple oopses 2.2.6 is unusable. In fact, I'll only use it to get a
> > working PPP when I need to download some more .debs.
>
> I have two days ago compiled 2.2.6, and everything worked. Where
> do you see the oopses happening?
In init (sometimes), login, klogd (really bad this one). Here is an
example oops (I used -c 4 as the -c 1 output looked dubious):
Unable to handle kernel paging request at virtual address fffffe00000056b8
sshd(190): Oops 0
pc = [<fffffc00003759d4>] ra = [<fffffe0000001fc0>] ps = 0000
r0 = 0000000000000041 r1 = 0000000120143061 r2 = 0000000120143061
r3 = 0000000000000041 r4 = 0000000000000041 r5 = 0000000000000000
r6 = fffffc0001fa9b60 r7 = 0000000120143020 r8 = fffffc0001668000
r9 = fffffc000146e120 r10= fffffc0001fa9b60 r11= 0000000000000041
r12= 0000000000000000 r13= fffffc0001fabb40 r14= fffffc000166beb8
r15= fffffc000166bdd8
r16= fffffc0001fa9b60 r17= fffffc000166bf00 r18= 0000000000000041
r19= 0000000000000015 r20= fffffc000166bd98 r21= 0000000000000000
r22= 0000000000000041 r23= fffffc000166bef8 r24= fffffffffffffff2
r25= fffffc000166c000 r27= fffffc0000375980 r28= 0000000000000001
gp = fffffe000000baf0 sp = fffffc000166bd58
Code: 46d60400 f4a00003 46100406 <a77d9bc8> 6b9b4009 f400000c a4570000 42560132 42160410
Trace; fffffc0000371324 <sock_sendmsg+a4/100>
Trace; fffffc00003715ec <sock_write+8c/a0>
Trace; fffffc000033ca7c <sys_write+13c/1a0>
Trace; fffffc0000310c20 <entSys+a8/c0>
Trace; fffffc0000310b78 <entSys+0/c0>
>>PC ; fffffc00003759d4 <memcpy_fromiovec+54/a0>
Code; fffffc00003759c8 <memcpy_fromiovec+48/a0> 0000000000000000 <_PC>:
Code; fffffc00003759c8 <memcpy_fromiovec+48/a0> 0: 00 04 d6 46 mov t8,v0
Code; fffffc00003759cc <memcpy_fromiovec+4c/a0> 4: 03 00 a0 f4 bne t4,14 <_PC+0x14> fffffc00003759dc <memcpy_fromiovec+5c/a0>
Code; fffffc00003759d0 <memcpy_fromiovec+50/a0> 8: 06 04 10 46 mov a0,t5
Code; fffffc00003759d4 <memcpy_fromiovec+54/a0> c: c8 9b 7d a7 ldq t12,-25656(gp)
Code; fffffc00003759d8 <memcpy_fromiovec+58/a0> 10: 09 40 9b 6b jsr at,(t12),38 <_PC+0x38> fffffc0000375a00 <memcpy_fromiovec+80/a0>
Code; fffffc00003759dc <memcpy_fromiovec+5c/a0> 14: 0c 00 00 f4 bne v0,48 <_PC+0x48> fffffc0000375a10 <memcpy_fromiovec+90/a0>
Code; fffffc00003759e0 <memcpy_fromiovec+60/a0> 18: 00 00 57 a4 ldq t1,0(t9)
Code; fffffc00003759e4 <memcpy_fromiovec+64/a0> 1c: 32 01 56 42 subl a2,t8,a2
Code; fffffc00003759e8 <memcpy_fromiovec+68/a0> 20: 10 04 16 42 addq a0,t8,a0
Trace: [<fffffc0000371324>] [<fffffc0000372348>] [<fffffc00003723bc>] [<fffffc0000310c20>] [<fffffc0000310b78>]
Does anyone recognise this? All the others seem to be in sock_sendmsg
and memcpy_fromiovec as well.
> Right now the only issue I have is with
> Plug and Pray in the kernel. It is really sparsely documented, and my
> soundcard still needs isapnptools to configure. But after the despair I
> had in upgrading I am very happy right now.
Plug and play is always fun the first few times. However, I find editing
the output from pnpdump is really easy, if I always modify copies of
the configuration lines rather than the original lines themselves.
> The transition from 1.2 to 2.0
> was about as painful, but at least I could install from scratch. Not an
> option with debian-alpha yet since there aren't any bootdisks out yet. But
> that's just one of the dangers of living on the edge.
The slink CD ROM has a Jensen kernel, but it is quite hard (the first 2/3
times) to get to a state with a decently partitioned HD and a bootable
kernel on the disk. It is really annoying that the various ideas that
fdisk has about cylinders (in BSD slice mode) do not match!
As an aside, I was not able to get Ethernet working properly either,
but I'm not sure where the fault lies (hint: the cable transports packets
in one direction only :).
One last thing, with 2.2.6-ac1 (and 2.2.7-pre1), I get a forkbomb of
modprobe processes at start up (very slow start up, something-scroll-lock
reveals hundreds of modprobes which eventually die or can be killed off
with ctrl-c). I got rid of auto in /etc/modules and this seems to have
stopped this.
Giuliano.
Reply to: