> 6) At firewall startup, BSOD
> 7) Subsequent restarts result in BSOD as previously stated.
Xen appears to be sending 'large' packets onto the bridge, and the tap
interface isn't 'un-large-ing' them.
"
20:40:34.437312 IP (tos 0x0, ttl 128, id 34470, offset 0, flags [DF],
proto: TCP (6), length: 52) 192.168.207.200.5001 >
192.168.207.254.53981: ., cksum 0x05aa (correct), 1:1(0) ack 14505 win
58295 <nop,nop,timestamp 57828874 1445680997>
20:40:34.437324 IP (tos 0x0, ttl 64, id 51510, offset 0, flags [DF],
proto: TCP (6), length: 4396) 192.168.207.254.53981 >
192.168.207.200.5001: . 23193:27537(4344) ack 1 win 46
<nop,nop,timestamp 1445680998 57828874>
20:40:34.439653 IP (tos 0x0, ttl 128, id 34471, offset 0, flags [DF],
proto: TCP (6), length: 64) 192.168.207.200.5001 >
192.168.207.254.53981: ., cksum 0x2a79 (correct), 1:1(0) ack 15953 win
56847 <nop,nop,timestamp 57828874 1445680997,nop,nop,sack 1
{17401:18849}>
20:40:34.439670 IP (tos 0x0, ttl 64, id 51513, offset 0, flags [DF],
proto: TCP (6), length: 4396) 192.168.207.254.53981 >
192.168.207.200.5001: . 27537:31881(4344) ack 1 win 46
<nop,nop,timestamp 1445680999 57828874>
"
Note the lengh of 4396 - the hardware interface is supposed to break
that up into MSS sized chunks, but there is no hardware interface
involved in this case.
I'm trying to resolve the problem in the GPL PV drivers.
I've had wireshark (npf.sys) BSOD when it receives a packet with a size
greater than MTU, so I imagine a firewall may well do the same thing.
Try turning off large send offload for all interfaces on the bridge (the
same bridge as your crashing DomU), eg:
"
ethtool -K <ifname> gso off
"
Iperf gives horrible results too when DomU is the 'server' and Dom0 is
the 'client' because of this.
James
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
Home