On Monday 21 Dec 2015 23:49:02 Kai Krakow wrote: > Am Mon, 21 Dec 2015 23:20:24 +0000 > > schrieb Mick <michaelkintz...@gmail.com>: > > On Monday 21 Dec 2015 22:38:21 Alan McKinnon wrote: > > > On 22/12/2015 00:37, Mick wrote: > > > > Am I alone in experiencing this? Any ideas for fixing it? > > > > Should I post a bug and if so where? I'm thinking that asking > > > > Lennart to fix ifplugd so that it works with openrc would be > > > > taking it a step too far. :p > > > > > > read the news items for the last 3 months or so. > > > > > > openrc changes, networking and service dependencies have changed > > > somewhat, all well documented in the news items > > > > Thank you Alan, I believe I've read the news item you refer to, but I > > am not sure I understand how or why it applies to two laptops. The > > news item refers to localmount and netmount which if they cannot be > > mounted will cause problems. I don't use netmount on these laptops: > > > > # rc-update -s -v | grep mount > > > > localmount | boot > > > > mount-ro | shutdown > > netmount | > > > > localmount works and I see no failures in mounting my filesystems. > > What fails repeatably is ifplugd, which no longer starts wired or > > wireless devices. > > You could try running the service in debug mode (--debug), and look at > the line starting ifplugd. Check if the parameters look correct, then > try to fire the same command from command line and check the console > output for errors.
ifplugd does not have a debug mode, but starting it by hand I see this: Dec 22 00:38:08 dell_xps ifplugd(eth0)[4982]: ifplugd 0.28 initializing. Dec 22 00:38:08 dell_xps ifplugd(eth0)[4982]: Using interface eth0 Dec 22 00:38:08 dell_xps ifplugd(eth0)[4982]: Failed to detect plug status of eth0 Dec 22 00:38:08 dell_xps ifplugd(eth0)[4982]: Exiting. Ha! It now assumes that I have an eth0 interface, which I don't. Hmm ... I wonder where it reads this from. If I start the NIC manually I get: Dec 22 00:33:52 dell_xps kernel: IPv6: ADDRCONF(NETDEV_CHANGE): enp11s0: link becomes ready Dec 22 00:35:31 dell_xps kernel: IPv6: ADDRCONF(NETDEV_UP): enp11s0: link is not ready Dec 22 00:35:30 dell_xps ifplugd(enp11s0)[4293]: ifplugd 0.28 initializing. Dec 22 00:35:30 dell_xps ifplugd(enp11s0)[4293]: Using interface enp11s0/00:26:B9:20:B4:9C with driver <tg3> (version: 3.137) Dec 22 00:35:30 dell_xps ifplugd(enp11s0)[4293]: Using detection mode: SIOCETHTOOL > By the way: Is ifplugd still needed? I thought it has been replacable > by openrcs improved hotplug/coldplug support a long time ago? I > remember my main purpose of using ifplugd back the days I still used it > was to decouple the network init latency from the rest of the boot > process and thus speed it up - in addition to the benefit of not > starting some services at all when no network was plugged in. Yes, ifplugd is needed here. Otherwise removal/insertion of the ethernet cable does not bring up/down the enp11s0 interface. > Part of the problem may be that udev handles device renaming and > bootstrapping now. Maybe ifplugd should be called from udev and no > longer be a service? In the end, ifplugd is tied to the device name - > and thus must be part of the device service it is going to manage. And > I don't think this is how it works in openrc nowadays. Thus it needs > to move deeper down the layer - which is udev. Thank you for pointing me in this direction. If ifplud now thinks that eth0 is the default interface, where might it be deducing this from? I installed sys-fs/udev-225 and dev-libs/libgudev-230 at the same time I updated openrc and this is when the problems started. I never had an ifplugd.conf in use, perhaps I ought to set up one now? > Alternatively, NetworkManager is probably also a good replacement then - > tho I totally understand why you wouldn't want to install it. At least > you wouldn't have to ask Lennart then for fixing ifplugd in openrc > if problems arise. ;-) > > As a side note: systemd-networkd handles and fits the same purpose as > ifplugd very well - should you consider to migrate to systemd. *scnr* > > It's also not by Lennart, which elevates your problems with asking > Lennart to fix something. :-) Thank you for this suggestion, but for reasons covered exhaustively and exhaustingly in a few past mega-threads I'd rather stay with openrc. -- Regards, Mick
signature.asc
Description: This is a digitally signed message part.
