SMP kernel with nosmp option problem

Zdenek Kabelac (kabi@fi.muni.cz)
Tue, 7 Dec 1999 20:01:54 GMT


Hello

For a couple of week I'm having this problem:

I've tried to use SMP kernel with nosmp option for kernel image.
And I have found out strange problem.
After a discussions with AC and testing few more cases
these are results.

Could anyone test this and confirm that its not just mine
computer?

Se here are the results:

Computer is 2x400 (450) Celer BP6 128MB G400 HD20GB
I'm using /usr & /home partition mounted over network
from another computer.
This machine also exports its directories (using knfs).

Everything works without almost any problems while normaly using SMP.
(I think once this computer locked when there was huge disk activity
on the nfs server)
However when I pass 'nosmp' parameter I'm getting into two states:

One of them is this:
Nov 30 16:53:37 dual kernel: eth0: 3Com 3c905C Tornado at 0xb000,
***INVALID CH
Nov 30 16:53:37 dual kernel: 8K byte-wide RAM 5:3 Rx:Tx split,
autoselect/Auto
Nov 30 16:53:37 dual kernel: Media override to transceiver type 4
(100baseTX).
Nov 30 16:53:37 dual kernel: Enabling bus-master transmits and
whole-frame rec
Nov 30 16:53:48 dual kernel: nfs: server XXXXXX not responding, still
try
Nov 30 16:53:48 dual kernel: nfs: server XXXXXX not responding, still
try
Nov 30 16:54:07 dual kernel: eth0: transmit timed out, tx_status 00
status e601.
Nov 30 16:54:07 dual kernel: eth0: Interrupt posted but not delivered --
IRQ blo
Nov 30 16:54:07 dual kernel: Flags; bus-master 1, full 0; dirty 433
current 43
Nov 30 16:54:07 dual kernel: Transmit list 00000000 vs.
c7343210.
Nov 30 16:54:07 dual kernel: 0: @c7343200 length 80000096 status
80010096
Nov 30 16:54:07 dual kernel: 1: @c7343210 length 8000009a status
0001009a
....
Nov 30 16:54:07 dual kernel: 14: @c73432e0 length 80000096 status
00010096
Nov 30 16:54:07 dual kernel: 15: @c73432f0 length 8000009a status
8001009a
Nov 30 16:54:07 dual kernel: eth0: Resetting the Tx ring
pointer.
Nov 30 16:54:07 dual kernel: nfs: server XXXX OK
Nov 30 16:54:07 dual kernel: nfs: server XXXX OK
Nov 30 16:54:17 dual kernel: nfs: server XXXX not responding, still try
Nov 30 16:54:19 dual kernel: nfs: task 416 can't get a request
slot
Nov 30 16:54:19 dual kernel: nfs: task 418 can't get a request
slot
Nov 30 16:54:19 dual kernel: nfs: task 417 can't get a request
slot

In this state I could use Magic key to remount drives RO and safely
reboot.
The second version is not so nice - as I'm getting messages about
hda: interrupt losts

In both cases there is nothing wrong when I stay in init mode 1.
In this mode when I print /proc/interrupts there are no conflicts on
irq.

As soon as I start nfs services this bug appears.

I have tried two other cards - EEpro100 and a different 3c59x card
(Bummerang).

I suspect probably some bug in nfs code, but I'm not sure about this.
Also I have tried many kernels - 2.2.13....2.2.14pre11 and also
2.2.30pre3
and all of them have the same problem on my computer.

When I've tried to run 2.2.13 with idk and nosmp I could not boot at
all.

Doesn't anyone have any ideas where could be the problem?

-- 
         binary drivers cause sclerosis of the kernel (mark hahn)
  Zdenek Kabelac  http://i.am/kabi/ kabi@i.am {debian.org; fi.muni.cz}

- 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/