| View previous topic :: View next topic |
| Author |
Message |
Jewelie
n00b
Joined: 16 Nov 2003
Posts: 6
Location: UK
|
 Posted: Thu Jun 10, 2004 1:57 pm Post subject: Inconsistent behaviour: pcmcia network vs dhcp |
 |
|
How do,
I've got a persistant issue that we've had since we've been using gentoo with the laptop since last year, and although we've been through various package revisions I can't for the life of me eliminate it.
I've got a network card, installed in a pcmcia socket at boot time, fully autoconfiguring via dhcp upon boot. It's important that it is as automatic as possible (as it'll be in different environments, some with networking, some without, and lots of different setups.)
What's odd is that I'm get different card manager related boot messages on different occasions and I can't work out why.
Here is one variation...
| Quote: |
* Starting pcmcia...
cardmgr[8658]: watching 2 sockets
cardmgr[8658]: starting, version is 3.2.5
cardmgr[8658]: socket 1: 3Com 572/574 Fast Ethernet
cardmgr[8658]: executing: 'modprobe 3c574_cs'
cardmgr[8658]: executing: './network start eth0'
cardmgr[8658]: + * Bringing eth0 up via DHCP... [ ok ]
cardmgr[8658]: + * eth0 received address
|
And here is the, rather less healthy sounding, other one...
| Quote: |
* Starting pcmcia...
cardmgr[8658]: watching 2 sockets
cardmgr[8658]: starting, version is 3.2.5
cardmgr[8658]: socket 1: 3Com 572/574 Fast Ethernet
cardmgr[8658]: executing: 'modprobe 3c574_cs'
cardmgr[8658]: executing: './network start eth0'
cardmgr[8658]: + * WARNING: "net.eth0" has already been started. [ ok ]
|
What exactly is starting eth0? Why should the behaviour be inconsistant.
After boot, the network appears to autoconfigure and behave properly in both situations, and /etc/resolv.conf is suitably filled in, and so on. However, I'm getting a matching cardmgr complaints on shutdown about being unable to stop eth0 neatly 'cos it already has been stopped.
My suspicion is that the differences are caused by a timing issue, and that there's two programs/methods which are attempting to start and stop the interface. I'd love to know what is starting net.eth0 before the pcmcia script, and I'd love to clean this up a bit.
Our rc-update -s
| Code: |
acpid |
apmd | default
bootmisc | boot
bootsplash |
checkfs | boot
checkroot | boot
clock | boot
consolefont | boot
crypto-loop |
cupsd |
domainname |
esound |
fam |
famd |
hdparm |
hostname | boot
hotplug | default
keymaps | boot
lisa |
local | default nonetwork
localmount | boot
modules | boot
mysql |
nessusd |
net.eth0 |
net.lo | boot
net.ppp0 |
netmount | default
nfs | default
nfsmount |
nscd |
ntp-client |
ntpd | default
numlock |
pcmcia | boot
portmap | default
postgresql |
reslisa |
rmnologin | boot
rsyncd |
samba |
serial | boot
sshd | default
switch |
syslog-ng | default
urandom | boot
vcron | default
winbind |
xdm | default
xfs | default
xinetd |
|
Our /etc/conf.d/net (only line not commented out)-
Our /etc/conf.d/pcmcia-
| Quote: |
# Put cardmgr options here
CARDMGR_OPTS="-f"
# To set the PCMCIA scheme at startup...
SCHEME="home"
# If using kernel PCMCIA drivers, PCIC should be "yenta_socket". If
# using the pcmcia-cs drivers, this shhould be either "i82365" or "tcic",
# depending on your pcmcia hardware.
# If using kernel drivers not as modules, set PCIC to ""
PCIC="i82365"
# Put socket driver timing parameters here
PCIC_OPTS=""
# Alternative PCIC driver to use if PCIC driver fails
PCIC_ALT="yenta_socket"
PCIC_ALT_OPTS=""
# Put pcmcia_core options here
CORE_OPTS=""
|
And lastly, the package versions installed are
sys-kernel/gentoo-sources-2.4.25-r3 (recompiled, and not using the kernel PCMCIA drivers)
sys-kernel/genkernel-3.0.2a
sys-apps/baselayout-1.9.4-r2
sys-apps/pcmcia-cs-3.2.5-r1 (using these pcmcia drivers)
sys-apps/hotplug-20040401
net-misc/dhcpcd-1.3.22_p4-r2
Any ideas what I'm doing wrong with this setup, or how I could clean it up (nicely, without forcing or hacking.)
Ta-ra,
Julie
PS Just for reference, etc & env-update have been run, and I'm not a Linux or computing or networking newbie, but I am admittedly a PCMCIA newbie.
_________________
Julie Brandon
http://www.computergeeks.co.uk/ |
|
| Back to top |
|
 |
dsd
Developer
Joined: 30 Mar 2003
Posts: 2162
Location: nr London
|
| Posted: Sat Jun 12, 2004 9:54 pm Post subject: |
|
|
it may well be hotplug racing against cardmgr. are you able to temporarily "rc-update del hotplug" and still have a bootable system?
it would be worth checking like this...
_________________
http://dev.gentoo.org/~dsd |
|
| Back to top |
|
|
Jewelie
n00b
Joined: 16 Nov 2003
Posts: 6
Location: UK
|
| Posted: Mon Jun 14, 2004 5:38 am Post subject: |
|
|
| dsd wrote: | it may well be hotplug racing against cardmgr. are you able to temporarily "rc-update del hotplug" and still have a bootable system?
it would be worth checking like this... |
Many thanks for the reply.
Your suggesting is such a ruddy logical and obvious thing to try I am dreadfully embarrassed I didn't do so myself (I don't have anything that ought to be relying on hotplug.) 
Unfortunately, it doesn't make any difference it seems. Bummer! Still get a random choice of those two startup messages, and always this on shutdown-
| Code: |
* Stopping pcmcia...
cardmgr[11229]: executing: './network check eth0'
cardmgr[11229]: executing: './network stop eth0'
cardmgr[11229]: + * ERROR: "net.eth0" has not yet been started.
cardmgr[11229]: executing: 'modprobe -r 3c574_cs'
cardmgr[11229]: exiting
|
However, if bring pcmcia down manually during normal system operation (i.e. outside of shutdown) I get this in the logs...
| Code: |
Jun 14 06:59:01 laptop rc-scripts: WARNING: you are stopping a boot service.
Jun 14 06:59:01 laptop cardmgr[10939]: executing: './network check eth0'
Jun 14 06:59:01 laptop cardmgr[10939]: executing: './network stop eth0'
Jun 14 06:59:02 laptop cardmgr[10939]: + * Bringing eth0 down
Jun 14 06:59:02 laptop cardmgr[10939]: + * Removing inet6 addresses
Jun 14 06:59:02 laptop cardmgr[10939]: + * eth0 inet6 del fe80::250:4ff:fe8e:9b9b/64.
.. [ ok ]
Jun 14 06:59:03 laptop dhcpcd[10942]: terminating on signal 10
Jun 14 06:59:03 laptop cardmgr[10939]: + * Releasing DHCP lease for eth0... [ ok ]
Jun 14 06:59:03 laptop cardmgr[10939]: + * Stopping eth0... [ ok ]
Jun 14 06:59:03 laptop cardmgr[10939]: executing: 'modprobe -r 3c574_cs'
Jun 14 06:59:03 laptop divert: freeing divert_blk for eth0
Jun 14 06:59:03 laptop cardmgr[10939]: exiting
Jun 14 06:59:03 laptop unloading PCMCIA Card Services
|
This too me looks cleaner and probably the way in which it should be coming down? Perhaps its the point at which the shutdown & startup is being performed in relation to something else? Could it be some odd interaction between dhcpcd & pcmcia? pcmcia is one of the last things to be stopped on shutdown, so could dhcpcd be being killed off before pcmcia is stopped, or some other "order in which things are stopped" be going wrong?
Worth noting that previous searches of the forums suggests I'm not the only person see this (in fact I'd be interested to know if anyone else doing automatic configuration of a PCMCIA network at boot via DHCP [isn't seeing the same thing!) However, as it doesn't seem to have any other yet noticed side-effects I don't think anyone has bothered to track it down (I'm the kind of obsessive compulsive person that at least likes to know what's generating an error message and why before I'm comfortable deciding to dismiss it!)
By the way, do you know if dhcpcd should be setting the hostname for the box if it can (the dhcp server is about as populated with details as it is possible for a dhcp server to be, however, ironically, I can't tell from the dhcpd man page whether or not out of all the info it DHCP actually INFORMS a client what its 'net hostname should actually be)?
Ta-ra,
Julie
_________________
Julie Brandon
http://www.computergeeks.co.uk/ |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
Networking & Security
