Re: NetBSD/xen goes off the network - reproduceable

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Tue, Feb 21, 2012 at 02:28:08PM -0500, Brian Marcotte wrote:
> > the attached patch changes xennet(4) to use a pool for receive
> > buffer, and allocate them on demand for the ring. If a receive buffer
> > is stalled in some socket queue, it's should not be a problem any more
> > (other than memory consumption).
> 
> I applied the patch to NetBSD-6 and I can no longer provoke the problem
> with my simple test.
> 
> Thanks!
> 
> However, running "tcpdump -i xennet0 -n icmp" reliably causes the panic
> below after 1-15 seconds. There may be one other issue, but I'm going to
> need another day to look into it. I think it may be something I'm doing.
> 
> 
> fatal page fault in supervisor mode
> trap type 6 code 0 eip c0108e36 cs 9 eflags 10287 cr2 506ff31 ilevel 8
> kernel: supervisor trap page fault, code=0
> Stopped in pid 0.2 (system) at  netbsd:m_xhalf+0x76:    movzwl  0(%edi),%esi
> m_xhalf(c9ff3e38,c02c3c54,c0b08154,c1720900,4ce14000,4,c040f7c2,ee6b4000,c9ff3e38,c0ad1e7c)
>  at netbsd:m_xhalf+0x76
> bpf_filter(c16465b4,c0bd6000,0,0,1d00,100,ab1000,0,c03ab1c0,0) at 
> netbsd:bpf_filter+0x2e0
> _bpf_mtap(c0aba724,c1720900,c0ad02a4,c0409160,c9ff3eb0,c04091c4,c04091c0,0,4f43e1a5,c0acf018)
>  at netbsd:_bpf_mtap+0xb0
> xennet_handler(c0acf000,0,c0400f80,c065b000,e,c0aba4e0,c9ff3fe8,c013d15b,c0aba4e0,ca18ac80)
>  at netbsd:xennet_handler+0x3d8
> intr_biglock_wrapper(c0aba4e0,ca18ac80,0,c9ff3fec,c016aaf7,e,1,6,16a80,c055c458)
>  at netbsd:intr_biglock_wrapper+0x1f
> evtchn_do_event(e,ca18ac80,ca18abf8,0,0,0,0,0,0,0) at 
> netbsd:evtchn_do_event+0x16b

i see it too; I'm looking at this.

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--