Maybe it is a bug of L4/Fiasco or L4Linux

[Adam Lackorzynski]

Hi,

On Mon Sep 18, 2006 at 11:19:27 +0800, chenggh wrote:
> 	When I use netpipe to test the performance of L4Linux based on top of 
> L4/Fiasco there is something as following:
> 
> kernel BUG at include/linux/netdevice.h:886!
> Trap: 6: 0000 [#1]
> Modules linked in:
> CPU:    0
> EIP:    f032:[<0056f317>]    Not tainted VLI
> EFLAGS: 00013246   (2.6.17-l4 #1)
> EIP is at rtl8139_poll+0xd7/0xf0
> eax: 00000006   ebx: 0001d43c   ecx: 0000c07f   edx: 0000d43c
> esi: 01331400   edi: b09fff10   ebp: b09fff00   esp: b09ffedc
> ds: 4000   es: ea0c   ss: 0000
> Process swapper (pid: 0, threadinfo=b09fe000 task=0066b940)
> Stack: 01331400 01331660 00000040 00000001 0001d400 01331660 013314c0 01331400
>        00003e13 b09fff20 005a8262 01331400 b09fff10 0000012c 00000001 007207d8
>        0000000a b09fff38 0041a7a9 007207d8 00000000 0000000c 00000000 b09fff48
> Call Trace:
>  <0040bec0> show_stack_log_lvl+0x90/0xc0  <0040c08b> 
> show_registers+0x14b/0x1c0
>  <0040c2bd> die+0xfd/0x200  <00000000> _start+0xffc00000/0x50
>  <005a8262> net_rx_action+0x72/0x110  <0041a7a9> __do_softirq+0x49/0xb0
>  <00407bc8> do_softirq+0x48/0x50  <0041a8c5> irq_exit+0x35/0x40
>  <00407993> do_IRQ+0x23/0x30  <0050b052> irq_thread_hw+0x1f2/0x330
>  <a00a8a4c> 0xa00a8a4c  <a00abc76> 0xa00abc76
> Code: 41 04 00 02 20 00 c7 86 c0 00 00 00 00 01 10 00 0f ba 76 2c 05 e8 4a 4a 
> e9 ff 31 c0 83 7d e8 00 0f 94 c0 83 c4 18 5b 5e 5f c9 c3 <0f> 0
> b 76 03 e0 03 64 00 eb b6 eb 0d 90 90 90 90 90 90 90 90 90
> EIP: [<0056f317>] rtl8139_poll+0xd7/0xf0 SS:ESP 0023:b09ffed4
>  <0>Kernel panic - not syncing: Fatal exception in interrupt
>  panic: going to sleep forever, bye
> NETDEV WATCHDOG: eth1: transmit timed out
> eth1: Transmit timeout, status 0c 0005 c07f media 10.
> eth1: Tx queue start entry 107084  dirty entry 107080.
> eth1:  Tx descriptor 0 is 0008a042. (queue head)
> eth1:  Tx descriptor 1 is 0008a042.
> eth1:  Tx descriptor 2 is 0008a5ea.
> eth1:  Tx descriptor 3 is 0008a5ea.
> eth1: link up, 100Mbps, full-duplex, lpa 0x41E1
> NETDEV WATCHDOG: eth1: transmit timed out
> eth1: Transmit timeout, status 0c 0005 c07f media 10.
> eth1: Tx queue start entry 4  dirty entry 0.
> eth1:  Tx descriptor 0 is 0008a4ef. (queue head)
> eth1:  Tx descriptor 1 is 0008a5ea.
> eth1:  Tx descriptor 2 is 0008a5ea.
> eth1:  Tx descriptor 3 is 0008a5ea.
> eth1: link up, 100Mbps, full-duplex, lpa 0x41E1
> NETDEV WATCHDOG: eth1: transmit timed out
> eth1: Transmit timeout, status 0c 0005 c07f media 10.
> eth1: Tx queue start entry 4  dirty entry 0.
> eth1:  Tx descriptor 0 is 0008a5ea. (queue head)
> eth1:  Tx descriptor 1 is 0008a5ea.
> eth1:  Tx descriptor 2 is 0008a5ea.
> eth1:  Tx descriptor 3 is 0008a5ea.
> eth1: link up, 100Mbps, full-duplex, lpa 0x41E1
> NETDEV WATCHDOG: eth1: transmit timed out
> eth1: Transmit timeout, status 0c 0005 c07f media 10.
> eth1: Tx queue start entry 4  dirty entry 0.
> eth1:  Tx descriptor 0 is 0008a5ea. (queue head)
> eth1:  Tx descriptor 1 is 0008a03c.
> eth1:  Tx descriptor 2 is 0008a03c.
> eth1:  Tx descriptor 3 is 0008a03c.
> eth1: link up, 100Mbps, full-duplex, lpa 0x41E1
> NETDEV WATCHDOG: eth1: transmit timed out
> 
> 	And every time the netpipe program don't exit normally and the L4Linux is 
> dead so that I must reboot it again.  Maybe it is a bug but I am not sure.
> 	Thank you.

Hard to tell what's going wrong here, I guess I need to reproduce this
to say anything about this. This is some Realtek 100Mbit NIC, right? Do
you give any special options to NPtcp?
Are both sides L4Linux or just one, if the latter, which side?



Adam
-- 
Adam                 adam at os.inf.tu-dresden.de
  Lackorzynski         http://os.inf.tu-dresden.de/~adam/