Hi, as reported several times guests with virtio-net crash under heavy network load. There have been several patches submitted to the upstream kernel e.g:
virtio_net: do not reschedule rx refill forever virtio_net: fix oom handling on tx udp/tcp/..: use limited socket backlog but that did not finally solved the problem. Is is at least possible to crash a recent 2.6.38 kernel very easily. There is currently a discussion going on on launchpad: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/579276 I found out earlier that an old patch by Suse developers seems to fix the problem for good: http://kerneltrap.org/mailarchive/linux-netdev/2010/6/4/6278660 and recently tests seem to confirm this. Does anyone know why this patch did not went upstream or was it overseen? Peter
