Hide Forgot
Description of problem: When trying to setup netconsole on a domU the modprobe fails because the network interface doesn't support pulling. Version-Release number of selected component (if applicable): 2.6.18-194.26.1 How reproducible: always Steps to Reproduce: 1. configure /etc/sysconfig/netconsole 2. service netconsole start 3. observe dmesg and /var/log/messages Actual results: kernel: netconsole: eth0 doesn't support polling, aborting. Expected results: netconsole starts and sends logs to remote host. Additional info: It appears that oracle has patched their kernel to support this. Reference is orabug 6993043
Created attachment 499178 [details] "Add xen pv netconsole support [orabug 6993043] [bz 7258]" "linux-2.6.18-xen-netconsole-el5u6.patch" from http://oss.oracle.com/el5/SRPMS-updates/kernel-2.6.18-238.0.0.0.1.el5.src.rpm
I applied the patch and tried to test it. Unfortunately, it doesn't seem to work at all from the PV domU to a separate physical machine, even though I got netconsole working from dom0 to the same different physical machine. I flushed the entire iptables config in dom0 (policies are ACCEPT), tried with virtual and then shared physical device. netconsole does start in domU, but there's still no traffic due to the domU kernel messages. I can send UDP packets from the domU to the remote syslog server (my laptop) just fine. I verified the target MAC address on these packets with tcpdump, within the domU, and it does match the netconsole target MAC setting. So, either this patch doesn't work, or I don't know how to set up netconsole in a PV domU. I can't find any Oracle bug tracker that would return anything useful when I enter "6993043" or "7258". I hunted these together from mailing lists, and I even had to extract the patch from an src rpm (see comment 1). Thus not much justification for the patch is available to me. This is the smaller problem. The bigger problem is, as I realize now, Paolo has already seen this patch (squashed together with another Oracle patch) and judged it wrong. See bug 605711 comment 0 (note the "[bz 7258]" substring), bug 605711 comment 7, and attachment 425157 [details]. Considering this patch doesn't seem to help where it should, violates written guidelines, and was not submitted to / reviewed by upstream, I'm closing this bug as WONTFIX.