|
|
|
|
|
|
|
|
|
|
xen-devel
Re: [Xen-devel] Hotplug script not working
From: "Ewan Mellor" <ewan@xxxxxxxxxxxxx>
On Thu, Nov 10, 2005 at 11:43:02PM +1300, Greg Brackley wrote:
I don't understand how hotplug would work on FC4, as FC4 uses udev
(sysctl
kernel.hotplug -> "kernel.hotplug = /sbin/udevsend", and udevd is
running),
but there is a limitation that xen only uses udev for udev 059 and
greater
('check_hotplug' script); but FC4 reports `udev- V` -> 058. How are other
people getting this to work on FC4? Why does the check_hotplug script
require udev >= 059?
udev 059 introduced new rule syntax and semantics, which we use.
If you don't have udev 059, you need to use the older hotplug stuff. Do
you
have an /sbin/hotplug?
I have worked around this with a very simple set of rules [1]. This
immediately allowed me to get domU images running. I'm unsure as to what the
effect of discarding all the other parameters has had.
I notice that 8 vif0/veth interfaces are created in dom0. I believe
providing a parameter to the driver would overcome this (if it worried me).
What concerned me slightly was a spurious vif5.0 appearing in dom0 (before
ANY domU's had been created).
I am now back to working out why I can get ICMP traffic working, but getting
other traffic forwarding is eluding me. It seems to relate to my
VLAN/bridge/vif configuration. I've looked at ecn, MAC addresses,
arp_filter, rp_filter, log_martians, bridge stp settings, vlan configs,
ethtool settings, and googled late into the night. I would still be VERY
keen to hear from anyone on successes (or failures) with VLANs and xen.
Possible areas to look at would be great. Strange thing is, I have one
machine (with a single VLAN) working ok [but I'm reluctant to touch it,
because I don't know why it works when the other machines don't].
Greg :-)
[1] /etc/udev/rules.d/10-xen-backend.rules
KERNEL=="vif*"
KERNEL=="veth*"
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
|
|
|
|