[Bug 335486] New: network interface not set up during boot after upgrade from 10.2 to 10.3
https://bugzilla.novell.com/show_bug.cgi?id=335486 Summary: network interface not set up during boot after upgrade from 10.2 to 10.3 Product: openSUSE 10.3 Version: Final Platform: i686 OS/Version: openSUSE 10.3 Status: NEW Severity: Normal Priority: P5 - None Component: Network AssignedTo: bnc-team-screening@forge.provo.novell.com ReportedBy: walter.haidinger@gmx.at QAContact: qa@suse.de Found By: --- After upgrading from openSUSE 10.2 to 10.3, the /etc/init.d/network script fails to configure the network interface (traditional method, no Networkmanager). It has worked without problems under 10.2. The script times out waiting for the mandatory device eth1 now: eth1 device: Marvell Technology Group Ltd. 88E8001 Gigabit Ethernet Controller (rev 13). eth1 DHCP client NOT running eth1 is down And indeed, as ifconfig shows, the interface isn't up. However, after booting, 'rcnetwork start' or 'yast2 network' works. Unfortunately /var/log/messages isn't very helpful other than: sgke eth1: Link is up at 1000 Mbps, full duplex ADDRCONF(NETDEV_CHANGE): eth1: links becomes ready skge eth1: disabling interface Please note that there is no delay in the above syslog entries (i.e. same timestamp). The interface seems to be brought down right after it was brought up. Since the above does not provide much details, I'd appreciate for some hints to further diagnose the problem. Thanks! -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486#c1
--- Comment #1 from Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486#c2
--- Comment #2 from Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486#c3
--- Comment #3 from Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486#c4
--- Comment #4 from Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486
Matej Horvath
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c5
Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c6
--- Comment #6 from Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User zoz@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c7
Christian Zoz
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c8
Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486
User peter-mailbox@web.de added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c9
--- Comment #9 from Peter Küppers
# Now get all available interfaces drop lo and separate them into physical and # not physical. Then get AVAILABLE_IFACES sorted to shutdown the not physical # first. for a in $(type_filter `ls -A /sys/class/net/`); do test "$a" = lo && continue; test "$a" = sit0 && continue; test "$a" = bonding_masters && continue; test "${a#wifi}" != "$a" && continue case $a in eth*|ath*|wlan*|ra*) # Skip these which are too new, they will come via hotplug #Stempeln in rename_netiface #- am Anfang: virgin #- während dem Schleifen: looping #- am Ende: renamed #ifup bricht gleich ab, wenn kein service network #Wenns keinen Stempel gibt dann Stempeln unknown --> skip #Wenn Stempel virgin --> skip # looping --> skip # renamed --> set up #In Statusschleife, wenn mandatory devices gecheckt werden: # wenn status failed # STEMPEL == unknown && halbe Wartezeit vorbei -> ifup # == virgin/looping/renamed -> nix STAMPFILE=$STAMPFILE_STUB`cat /sys/class/net/$a/ifindex` if [ "$MODE" == onboot -a "$ACTION" == start ] ; then if [ -r "$STAMPFILE" ] ; then case "`cat $STAMPFILE`" in virgin|looping) continue ;; esac else echo unknown > $STAMPFILE continue fi fi ;; esac for b in $DIALUP_IFACES $TUNNEL_IFACES; do if [ "$a" = "$b" ] ; then NOT_PHYSICAL_IFACES="$NOT_PHYSICAL_IFACES $a" continue 2 fi done case $a in sit*) NOT_PHYSICAL_IFACES="$NOT_PHYSICAL_IFACES $a" continue 2 ;; esac PHYSICAL_IFACES="$PHYSICAL_IFACES $a" done <<<
In openSUSE 10.3 /etc/init.d/network is different up from line 472:
# Now get all available interfaces drop lo and separate them into physical and # not physical. Then get AVAILABLE_IFACES sorted to shutdown the not physical # first. # Interfaces may be renamed by udev after they are registered. In some cases # this may take some time. Therefore we check a 'renamed' flag if an interface # is ready to be set up. If an it is not ready now, it will be set up via # udev/ifup (because network is started now). We will just have to wait for it # later in the start section if it is considered mandatory (see next section). for a in $(type_filter `ls -A /sys/class/net/`); do case "`get_iface_type $a`" in eth|tr|wlan) STAMPFILE=$STAMPFILE_STUB`cat /sys/class/net/$a/ifindex` if [ "$MODE" == onboot -a "$ACTION" == start ] ; then if [ ! -e "$STAMPFILE" ] ; then continue # this leaves the for-loop! fi fi ;; lo|wlan_aux) continue ;; esac for b in $DIALUP_IFACES $TUNNEL_IFACES; do if [ "$a" = "$b" ] ; then NOT_PHYSICAL_IFACES="$NOT_PHYSICAL_IFACES $a" continue 2 fi done case $a in sit*) NOT_PHYSICAL_IFACES="$NOT_PHYSICAL_IFACES $a" continue 2 ;; esac PHYSICAL_IFACES="$PHYSICAL_IFACES $a" done <<< I marked the line with the supposed bug with a comment (this leaves the for-loop!), see above. Since the for-loop is stopped here, the variable PHYSICAL_IFACES has no value! And what's about "virgin and looping" from openSUSE 10.2?
I hope this gives a hint for a solution to the bug. Peter -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User peter-mailbox@web.de added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c10
Peter Küppers
https://bugzilla.novell.com/show_bug.cgi?id=335486
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c11
--- Comment #11 from Marius Tomaschewski
/var/log/boot.msg with network debugging enabled: (please note the "unknown option" error messages!)
The "unknown option rc ignored" is just debug output from ifup-wireless script without any relevance (it just informs about ignored rc option). Please update the sysconfig package to the most recent update package: http://download.opensuse.org/update/10.3/rpm/i586/sysconfig-0.70.2-4.2.i586r... http://download.opensuse.org/update/10.3/rpm/x86_64/sysconfig-0.70.2-4.2.x86... then, please verify the package using rpm -V sysconfig When rpm reports some modification/inconsistence like: # rpm -V sysconfig S.5....T /etc/init.d/network remove the reported files and install the package again. Verify the udev installation using "rpm -V udev" as well. # ls -l /etc/udev/rules.d/*net*.rules | cut -b 23- 450 9. Mai 08:17 /etc/udev/rules.d/70-persistent-net.rules 1518 21. Sep 2007 /etc/udev/rules.d/75-persistent-net-generator.rules 823 24. Apr 00:26 /etc/udev/rules.d/77-network.rules The /etc/udev/rules.d/70-persistent-net.rules is generated and contains mapping of the hardware to the interface name. Please verify, it reflect your hardware address (MAC) and the ifcfg-<interface> files in /etc/sysconfig/network. (In reply to comment #9 from Peter Küppers)
I marked the line with the supposed bug with a comment (this leaves the for-loop!), see above. Since the for-loop is stopped here, the variable PHYSICAL_IFACES has no value! And what's about "virgin and looping" from openSUSE 10.2?
I hope this gives a hint for a solution to the bug.
No, the scripts / rule files (in sysconfig and udev) are rewritten for 10.3 and are simplier in many places. On 10.3, the "ifcfg-<hardware-description>" (ifcfg-eth-id-00:01:02:8E:21) support is removed completely, we just use ifcfg-<interfacename>. The PHYSICAL_IFACES variable is empty, because the interface is not up until now and we wait for udev to load the modules. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c12
--- Comment #12 from Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c13
--- Comment #13 from Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c14
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c15
--- Comment #15 from Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c16
--- Comment #16 from Marius Tomaschewski
Unfortunately the sysconfig update to -4.2 did not work for me.
And what else is wrong with it?
Please note the following things regarding this bug: * It only shows during boot. Subsequent rcnetwork calls once logged in succeed. * It's probably caused by the assumption that _only_ udevd loads the modules. But what if the module is already loaded before boot.udev is run?
No, even I add the forcedeth driver to the initrd (INITRD_MODULES variable in /etc/sysconfig/kernel + mkinitrd) and the driver is loaded before udev starts, it works fine for me - see attached boot.msg. Please update correctly. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c17
--- Comment #17 from Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User peter-mailbox@web.de added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c18
--- Comment #18 from Peter Küppers
Waiting for mandatory devices: eth0 eth1 __NSC__ 18 17 16 14 13 11 10 8 7 6 4 3 1 0 eth0 device: Realtek Semiconductor Co., Ltd. RTL-8139/8139C/8139C+ (rev 10) eth0 is down failed eth0 interface could not be set up until now .. <<<
If I add the following (from opensuse 10.2) in the network script up from line 493 instead of the hard 'continue', the script works fine for me:
493: if [ ! -e "$STAMPFILE" ] ; then case "`cat $STAMPFILE`" in virgin|looping) continue ;; esac else echo unknown > $STAMPFILE continue fi <<<
It must have something to do with the not properly filled PHYSICAL_IFACES variable (or around this, e.g. with udev?). Cheers Peter -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c19
--- Comment #19 from Walter Haidinger
Unfortunately the sysconfig update to -4.2 did not work for me. And what else is wrong with it?
Sorry, a bit misleading. I meant that the update itself worked but this bug still persists. No problems with updates. Marc, Peter, please could you too do the following in order to help Marius debugging the script: Add the 3 lines set >> /var/log/rcnetwork.set exec >> /var/log/rcnetwork.log 2>&1 set -v; set -x at the top of your /etc/init.d/network and reboot. This will log the shell variables during boot and the (expanded) commands run. Marius should then have something to compare to his script from three different machines, all experiencing the bug. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User peter-mailbox@web.de added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c20
--- Comment #20 from Peter Küppers
https://bugzilla.novell.com/show_bug.cgi?id=335486
User peter-mailbox@web.de added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c21
--- Comment #21 from Peter Küppers
https://bugzilla.novell.com/show_bug.cgi?id=335486
User peter-mailbox@web.de added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c22
--- Comment #22 from Peter Küppers
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c23
--- Comment #23 from Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c24
--- Comment #24 from Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c26
--- Comment #26 from Marius Tomaschewski
All this in the hope it will provide some useful information.
Yes and no: ++ ls -d /etc/sysconfig/network/ifcfg-eth0 /etc/sysconfig/network/ifcfg-eth3~ /etc/sysconfig/network/ifcfg-eth5~ /etc/sysconfig/network/ifcfg-lo /etc/sysconfig/network/ifcfg-type-wlan /etc/sysconfig/network/ifcfg-wlan0 /etc/sysconfig/network/ifcfg-wlan0-bu~ /etc/sysconfig/network/ifcfg-wlan0~ [...] + MANDATORY_DEVICES=' eth0 type-wlan wlan0 __NSC__ ' This means, there is a problem, that the ifcfg-type-wlan is not converted -- the hwdesc2iface script is unable to handle it. We will see with Chistian (in NEEDINFO) if it can be converted somehow; at the moment, just move it away (rename to ifcfg.type-wlan) and delete the "ifcfg-*~" files. This should fix your problem Marc. Beside of the above obsolete ifcfg-<hwdescr> files: Not the /etc/init.d/network script causes the problems, but something with udev and ifup. So please provide log files as described bellow. At http://www.suse.de/~mt/openSUSE/10.3/, in 10.3-<arch> subdirectories, you'll find sysconfig RPMs with enabled extra debug. Please install them, reset your /var/log/messages with: bzip2 -9c < /var/log/messages \ > /var/log/messages-$(date +%Y%m%d).bz2 && \ cp /dev/null /var/log/messages and _reboot_ (no, really not a joke - I want to see what udev is doing). After the reboot, please collect files and create an archive using e.g.: mkdir /tmp/bug-355786-extradebug cp -a /dev/shm/sysconfig/* /tmp/bug-355786-extradebug/ cp -a /var/log/messages /var/log/boot.msg /tmp/bug-355786-extradebug/ cd /tmp/bug-355786-extradebug/ #### replace wireless keys / passwords / another secets with XXXXXX tar cvzf /tmp/bug-355786-extradebug.tgz * And attach the archive as (private) attachment to this bug. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c27
--- Comment #27 from Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User peter-mailbox@web.de added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c28
--- Comment #28 from Peter Küppers
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c29
--- Comment #29 from Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c30
Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c31
--- Comment #31 from Marius Tomaschewski
Marius,
I was finally was able to create the debug info as requested by comment #26. Since I do not know how to make a private attachment (does somebody?), I'll simply send it to you directly.
OK, thanks! In your and in Peters case, a "ifup eth0 -o hotplug" is never called (same for eth1). This means, you have a problem with udev rules. /etc/udev/rules.d # ls -l *net* -rw-r--r-- 1 root root 450 2008-05-26 14:59:09 70-persistent-net.rules -rw-r--r-- 1 root root 1518 2007-09-21 21:12:39 75-persistent-net-generator.rules -rw-r--r-- 1 root root 823 2008-05-22 12:39:27 77-network.rules There should be one rule for each physical network device, e.g.: /etc/udev/rules.d # grep -Ev "^#|^$" 70-persistent-net.rules SUBSYSTEM=="net", DRIVERS=="?*", ATTR{address}=="00:17:31:ca:a5:a5", NAME="eth0" SUBSYSTEM=="net", DRIVERS=="?*", ATTR{address}=="00:17:31:ca:a3:92", NAME="eth1" Please verify using "rpm -V udev" and "rpm -V syslog" that the both another rule files are not modified: # rpm -qf /etc/udev/rules.d/75-persistent-net-generator.rules udev-114-19 # rpm -qf /etc/udev/rules.d/77-network.rules sysconfig-0.70.2-4.6 The 77-network.rules file is responsible for "marking" the interface available by creating the $STAMPFILE: ==> 77-network.rules: [...] SUBSYSTEM=="net", ACTION=="add", RUN+="/sbin/ifup $env{INTERFACE} -o hotplug" [...] ==> /sbin/ifup: [...] if [ "$SCRIPTNAME" == ifup -a "$HOTPLUG" == yes ] ; then IFINDEX=/sys/$DEVPATH/ifindex if [ -r "$IFINDEX" ] ; then STAMPFILE=$STAMPFILE_STUB`cat $IFINDEX` echo renamed > $STAMPFILE fi fi [...] The DEVPATH variable is provided by udev and points to the path of the device, e.g. devices/pci0000:00/0000:00:10.0/net/eth0. The STAMPFILE is checked in the rcnetwork script -- see comment #4 and #9, your patches apply exactly to this place.. But the network script is not the reason of the problem - something is wrong with the udev network rules on your systems. This is the reason, why I asked to verify sysconfig + udev installation. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c32
--- Comment #32 from Marius Tomaschewski
Created an attachment (id=217913) --> (https://bugzilla.novell.com/attachment.cgi?id=217913) [details] debug info as requested, from Marc
Marius,
As suggested, I deleted the "ifcfg-*~" files.
You said: This should fix your problem Marc. But what problem?
That the network script waits for an "type-wlan" interface that will be never available - see also in comment #27.
I also renamed ifcfg-type-wlan. There is no need to convert this one, I don't need it.
Then remove it, but don't rename to "ifcfg-wlan1" or the network script will wait for "wlan1" interface.
I rebooted with the original network script. The network was not started.
Same problem as in Peters and Walters case - something is wrong with udev rule files. "ifup eth0 -o hotplug" is never called. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c33
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c34
--- Comment #34 from Walter Haidinger
ls -l *net* shows the 70-,75- und 77- files that you listed.
# grep -Ev "^#|^$" 70-persistent-net.rules ok, there are entries for each ethX-interface.
rpm -V ... udev is quiet (no changes), and syslog-ng (not syslog!) shows only /etc/syslog-ng/syslog-ng.conf which is ok (added custom entries).
rpm -qf ... Same packages as yours.
77-network.rules has exactly the required line.
/sys/devices/pci*/net/ethX entries exist.
Replying to comment #33:
rpm -V udev sysconfig ok, nothing is printed.
None of the 5 udev rules listed in comment #33 exist on my system. The obsoleted 30-... rule was even documented in the Release Notes, IIRC. So yes, I'm aware of that. Now, without anything obvious wrong in the udev setup, how can we debug this further? What are the possible reasons that the "add" action of 77-network.rules is _not_ triggered? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c35
Marius Tomaschewski
Hi Marius, thanks for your help!
Thanks for your reports!
rpm -V ... udev is quiet (no changes), and syslog-ng (not syslog!) shows only /etc/syslog-ng/syslog-ng.conf which is ok (added custom entries).
OK. I meant "rpm -V sysconfig" of course, but you verified it bellow ;-)
Replying to comment #33:
rpm -V udev sysconfig ok, nothing is printed.
OK.
None of the 5 udev rules listed in comment #33 exist on my system.
The obsoleted 30-... rule was even documented in the Release Notes, IIRC. So yes, I'm aware of that.
Now, without anything obvious wrong in the udev setup, how can we debug this further?
What are the possible reasons that the "add" action of 77-network.rules is _not_ triggered?
Yes, this is the question here. It mayby happens, because of your 2.6.25.4-vmhost32 kernel... Perhaps the forcedeth driver does not trigger this event without patch? Hmm... Marcs and Peters seem to use suse default kernels... We will ask the maintainer and find it out - reassigning to him. The rename rule (70-persistent-net.rules) is called / visible in your boot.msg file. <6>udev: renamed network interface eth1 to eth0 <6>udev: renamed network interface eth0_rename to eth1 In my case, I can see more log lines when rename happens (10.3): <6>eth0 renamed to eth0_rename <6>eth1 renamed to eth0 <6>udev: renamed network interface eth1 to eth0 <6>eth0_rename renamed to eth1 <6>udev: renamed network interface eth0_rename to eth1 I think, it makes sense to enable ulog debug mode; perhaps it is visible there. Please set udev_log="debug" in /etc/udev/udev.conf, reboot and provide the logs - should be in /var/log/boot.msg. Alternatively, you can also try to trigger it at runtime: rcnetwork stop rmmod forcedeth # network card driver udevcontrol log_priority=debug rcnetwork start -o boot & # in background, or modprobe on another modprobe forcedeth # terminal after rcnetwork start... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kasievers@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c36
Kay Sievers
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c37
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kasievers@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c38
Kay Sievers
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c39
--- Comment #39 from Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c40
--- Comment #40 from Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c41
--- Comment #41 from Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kasievers@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c42
--- Comment #42 from Kay Sievers
Q: Who sets environment variable DEVPATH? udev? Required in /sbin/ifup (see comment #31)
The kernel. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c43
--- Comment #43 from Walter Haidinger
Yes, this is the question here. It mayby happens, because of your 2.6.25.4-vmhost32 kernel... Perhaps the forcedeth driver does not trigger this event without patch?
Checked patches of kernel-default-2.6.22.17-0.1.nosrc.rpm to drivers/net/forcedeth.c. There are some, but either already applied in current 2.6.25.4 or not applicable, e.g. patches to printk(). Unless there are other patches to the networking stack, I doubt that the forcedeth driver is the problem. Marc, Peter, which network driver do you use? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c44
Marius Tomaschewski
Udevtest shows, it should run. How did you find out, it is not called?
I've created a sysconfig package (http://www.suse.de/~mt/openSUSE/10.3/, see comment 26), that enables "bash -vx" debugging in both, rcnetwork and ifup: /sbin/ifup: R_INTERNAL=1 # internal error, e.g. no config or missing scripts cd /etc/sysconfig/network || exit $R_INTERNAL test -f ./config && . ./config test -f scripts/functions && . scripts/functions || exit $R_INTERNAL ###### scripts/functions creates /dev/shm/sysconfig scripts/extradebug /etc/sysconfig/network/scripts/extradebug: SCRIPT=${0##*/} if test -d /dev/shm/sysconfig ; then exec 2> /dev/shm/sysconfig/exdeb.${SCRIPT}_$$.$PPID.${SEQNUM}_$1.$2 [...] set -vx fi In the logs provided in comment 28 and comment 29, ifup is never called with the "-o hotplug" option as udev is starting it in the rule. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User peter-mailbox@web.de added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c45
--- Comment #45 from Peter Küppers
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c46
--- Comment #46 from Marius Tomaschewski
In reply to comment #35:
Yes, this is the question here. It mayby happens, because of your 2.6.25.4-vmhost32 kernel... Perhaps the forcedeth driver does not trigger this event without patch?
Checked patches of kernel-default-2.6.22.17-0.1.nosrc.rpm to drivers/net/forcedeth.c. There are some, but either already applied in current 2.6.25.4 or not applicable, e.g. patches to printk(). Unless there are other patches to the networking stack, I doubt that the forcedeth driver is the problem.
Marc, Peter, which network driver do you use?
It does not depend on the driver [at least as shipped by suse]. Marc and me are using forcedeth from suse kernel too and it works fine for me but not for Marc. Peter is using r8169 (suse). -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c47
--- Comment #47 from Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c48
--- Comment #48 from Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486
User peter-mailbox@web.de added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c49
--- Comment #49 from Peter Küppers
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c50
--- Comment #50 from Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c51
--- Comment #51 from Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User enikesha@gmail.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c52
Innocenty Enikeew
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kasievers@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c53
Kay Sievers
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c54
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c55
Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c56
--- Comment #56 from Marius Tomaschewski
#for p in $(pidof udevd) ; do ls -l /proc/$p/exe ; done lrwxrwxrwx 1 root root 0 Oct 3 07:42 /proc/2558/exe -> /sbin/udevd* ^^^^^^^^^^^ Do you doubt that I don't have udevd running?
No. I wanted to verify a bug that we've had on 11.1 Beta, where the udevd from initrd was still running due bug in sysvinit. It would appear as "-> /sbin/udevd (deleted)"... (In reply to comment #52 from Innocenty Enikeew)
Hello,
I also have this problem and tried to investigate it myself. Putting echo's here and there I've found that during /etc/init.d/network starts in 'onboot' mode it expects udev to rename interface to that time and if this isn't the case ifup wouldn't not run for it. Network interfaces for which udev rule has run ifup it creates STAMPFILE in /dev/shm/sysconfig/ with 'renamed' in it. But if udev called this rule before rcnetwork had started ifup exits with "Service network not started and mode 'auto' -> skipping"". Ifup creates that STAMPFILE during on boot udev device detection, but when rcnetwork ran after it, somehow /dev/shm/sysconfig/ is empty! And makes rcnetwork wait for interface's detection.
I'll verify this. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kl@httgmbh.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c57
Kai Lappalainen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c58
Marius Tomaschewski
We have this same problem here with 2 from 4 machines which were upgraded from 10.2 to 10.3. All have static ip-addresses, different network cards, different architecture.
The difference during the upgrade was, that the two working machines were upgraded from "outside" by booting with a network-install-cd. The two machines showing the problems were upgraded from a running system by changing the repos to the 10.3 versions and then using the "factory upgrade"-tool in yast.
[...] Updates of a running system to a new distribution are AFAIK not supported and yast2 shows AFAIK at least a red warning. This the case, because there are several problems that may occur. One example: the conversion of ifcfg-eth-id-* to ifcfg-ethX needs an already updated udev with an already generated persistent-net rule 70 or a kernel using the new sysfs to work propelly. When the conversion happens while the old udev is running, the rule 70 it generates does not exists and the conversion using the old one may result in different persistent name. Using the old sysfs also does not work, because they differ significantly and on the new system a completely different modules may be in use. ... But let's take a look to the rules / config. Perhaps we'll find it. Can you attach a tgz from a working and a not working machine? "tar cvzf /tmp/machine1.tgz /etc/udev /etc/sysconfig/network" In /var/adm/backup/sysconfig are backups - please provide one of each machine that contains the old configuration (ifcfg-eth-<hwdesc> files). Of course, please copy the dirs somewhere first, review and replace any private data with XXXXX / example.com / dummy IPs / ... and attach it with a private flag set. And please also provide the output of "rpm -V udev sysconfig"? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c59
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c60
--- Comment #60 from Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kl@httgmbh.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c61
--- Comment #61 from Kai Lappalainen
ls -l /dev/shm/sysconfig/ total 20 -rw-r--r-- 1 root root 5 Dec 17 23:49 config-eth0 -rw-r--r-- 1 root root 3 Dec 17 23:49 config-lo -rw-r--r-- 1 root root 27 Dec 17 23:49 if-lo -rw-r--r-- 1 root root 7 Dec 17 23:49 ifup-lo -rw-r--r-- 1 root root 3 Dec 17 23:49 network -rw-r--r-- 1 root root 0 Dec 17 23:49 ready-lo drwxr-xr-x 2 root root 60 Dec 17 23:49 tmp
udevadm trigger --verbose --retry-failed --subsystem-match="net" (no output)
ls -l /dev/shm/sysconfig/ total 20 -rw-r--r-- 1 root root 5 Dec 17 23:49 config-eth0 -rw-r--r-- 1 root root 3 Dec 17 23:49 config-lo -rw-r--r-- 1 root root 27 Dec 17 23:49 if-lo -rw-r--r-- 1 root root 7 Dec 17 23:49 ifup-lo -rw-r--r-- 1 root root 3 Dec 17 23:49 network -rw-r--r-- 1 root root 0 Dec 17 23:49 ready-lo drwxr-xr-x 2 root root 60 Dec 17 23:49 tmp
udevadm trigger --verbose --subsystem-match="net" /devices/pci0000:00/0000:00:02.0/0000:06:00.0/0000:07:00.0/0000:08:00.0/0000:09:00.0/net/eth1 /devices/pci0000:00/0000:00:1c.0/0000:04:00.0/0000:05:00.0/net/eth0 /devices/virtual/net/lo
ls -l /dev/shm/sysconfig/ total 36 -rw-r--r-- 1 root root 5 Dec 17 23:54 config-eth0 -rw-r--r-- 1 root root 3 Dec 17 23:49 config-lo -rw-r--r-- 1 root root 29 Dec 17 23:54 if-eth0 -rw-r--r-- 1 root root 27 Dec 17 23:49 if-lo -rw-r--r-- 1 root root 7 Dec 17 23:54 ifup-eth0 -rw-r--r-- 1 root root 7 Dec 17 23:49 ifup-lo -rw-r--r-- 1 root root 3 Dec 17 23:49 network -rw-r--r-- 1 root root 8 Dec 17 23:54 new-stamp-2 -rw-r--r-- 1 root root 8 Dec 17 23:54 new-stamp-3 -rw-r--r-- 1 root root 0 Dec 17 23:54 ready-eth0 -rw-r--r-- 1 root root 0 Dec 17 23:54 ready-eth1 -rw-r--r-- 1 root root 0 Dec 17 23:49 ready-lo drwxr-xr-x 2 root root 60 Dec 17 23:54 tmp
(eth1 is not used/configured!) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kl@httgmbh.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c62
--- Comment #62 from Kai Lappalainen
But let's take a look to the rules / config. Perhaps we'll find it. Can you attach a tgz from a working and a not working machine?
"tar cvzf /tmp/machine1.tgz /etc/udev /etc/sysconfig/network"
machine1 is the not working machine, machine2 is the working machine.
In /var/adm/backup/sysconfig are backups - please provide one of each machine that contains the old configuration (ifcfg-eth-<hwdesc> files).
the working machine had no ifcfg-eth-<hwdesc> files before the upgrade, but an ifcfg-eth0 file in the oldest sysconfig-backup I've found. (?) It's the same on the second working machine. I'll attach the files anyway.
Of course, please copy the dirs somewhere first, review and replace any private data with XXXXX / example.com / dummy IPs / ... and attach it with a private flag set.
And please also provide the output of "rpm -V udev sysconfig"?
No output on both machines. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kl@httgmbh.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c63
--- Comment #63 from Kai Lappalainen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kl@httgmbh.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c64
--- Comment #64 from Kai Lappalainen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kl@httgmbh.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c65
Kai Lappalainen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c66
Marius Tomaschewski
(In reply to comment #60 from Marius Tomaschewski)
I have no udevadm, so I used udevtrigger
sure, sorry.
ls -l /dev/shm/sysconfig/ total 36 -rw-r--r-- 1 root root 5 Dec 17 23:54 config-eth0 -rw-r--r-- 1 root root 3 Dec 17 23:49 config-lo -rw-r--r-- 1 root root 29 Dec 17 23:54 if-eth0 -rw-r--r-- 1 root root 27 Dec 17 23:49 if-lo -rw-r--r-- 1 root root 7 Dec 17 23:54 ifup-eth0 -rw-r--r-- 1 root root 7 Dec 17 23:49 ifup-lo -rw-r--r-- 1 root root 3 Dec 17 23:49 network -rw-r--r-- 1 root root 8 Dec 17 23:54 new-stamp-2 -rw-r--r-- 1 root root 8 Dec 17 23:54 new-stamp-3 -rw-r--r-- 1 root root 0 Dec 17 23:54 ready-eth0 -rw-r--r-- 1 root root 0 Dec 17 23:54 ready-eth1 -rw-r--r-- 1 root root 0 Dec 17 23:49 ready-lo drwxr-xr-x 2 root root 60 Dec 17 23:54 tmp
(eth1 is not used/configured!)
Then "all is fine", the new-stamp-* are created after explicit udevtrigger run for both interfaces (correct, they exists) and because the network was already started (network file exists), also configured (eth0). -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kl@httgmbh.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c67
Kai Lappalainen
Now, you've provided:
machine1/new Yes, as attachment 260717 to comment 62.
In machine1/old archive are ifcfg-eth-id-* files - ok, but:
The new files machine1 would be most interesting ... Can you provide them? There is afaik no backup of udev rules, except of the rule 30 that sysconfig creates during conversion... It's not clear to me what's missing? Could you please explain?
The machine2/new/udev/rules.d/* looks like a fresh install, not like an update. There is no converted & disabled rule 30.
Sorry, I've checked our logs. It turned out, that this server was (inplace) upgraded at September, 13th 2007 from 10.2 to 10.3 *factory* before doing the upgrade to 10.3 final at October, 4th 2007. So I'm afraid this machine is not comparable - other than maybe there was a change between Sept. 13th and Oct. 4th which broke the conversion, because this machine works? The same is true for a second working machine here, which was also upgraded before to 10.3 factory on Sept. 13th and also works. In the /etc/sysconfig/network backup from Oct., 4th there is an empty file called "__convert_hwdesc_to_iface__". Other than that I see no significant difference to the one provided. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c68
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kl@httgmbh.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c69
--- Comment #69 from Kai Lappalainen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c70
--- Comment #70 from Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c71
--- Comment #71 from Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kl@httgmbh.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c72
Kai Lappalainen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User peter-mailbox@web.de added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c73
--- Comment #73 from Peter Küppers
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c74
--- Comment #74 from Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c75
Marius Tomaschewski
Hello,
sorry for the late answer
I've the same result:
my /etc/fstab contains: tmpfs /dev/shm tmfs size=1G 0 0
when I remove (uncomment) the line, yes bingo that's it: the problem is fixed! but I need the line...
No, you don't need it: /dev is already a tmpfs. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c76
--- Comment #76 from Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User fnm@osier-farm.org.uk added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c77
--- Comment #77 from Neil Murphy
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c79
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c80
Marius Tomaschewski
Ah, sounds like I was suffering from the /dev/shm problem too (bug #410367).
ATI tell you to mount a tmpfs at /dev/shm in order to make 3d support work in their fglrx drivers.
I've gave up on ATI's drivers some time ago which probably explains why I've not seen this bug for a while.
Well, ATI, VMWARE, Oracle. By default there are up to 2G in /dev/shm: LANG=C df -h /dev/shm Filesystem Size Used Avail Use% Mounted on udev 2.0G 264K 2.0G 1% /dev but some applications may need more. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c81
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c82
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c83
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kasievers@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c84
--- Comment #84 from Kay Sievers
Is is possible to specify the size of the /dev fs in /etc/fstab?
It may be possible by adding size=, but making it larger than the default (half the RAM size) just papers over some utterly broken applications, which should be fixed instead. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c85
--- Comment #85 from Marius Tomaschewski
(In reply to comment #83 from Marius Tomaschewski)
Is is possible to specify the size of the /dev fs in /etc/fstab?
It may be possible by adding size=, but making it larger than the default (half the RAM size) just papers over some utterly broken applications, which should be fixed instead.
Yes, half of RAM size (2G was wrong / my machine only :). Well, but when you've 8G RAM and run only e.g. a database it may be required to set it to e.g. 6G. IMO we have at least two choices: a) change the init / network scripts to not to use it b) make it adjustable (/etc/init.d/boot mounts /dev using "mount -n -t tmpfs -o mode=0755 udev /dev" -- I'm going to test if the size is used when I add /dev with a different size to /etc/fstab)... In case of a) -- which path we can use instead - /dev/.tmp? /var may be mounted on a separate disk so it can't be used too. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c86
--- Comment #86 from Marius Tomaschewski
a) change the init / network scripts to not to use it b) make it adjustable (/etc/init.d/boot mounts /dev using "mount -n -t tmpfs -o mode=0755 udev /dev" -- I'm going to test if the size is used when I add /dev with a different size to /etc/fstab)...
No, the size for /dev isn't used... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c87
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c88
--- Comment #88 from Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c89
--- Comment #89 from Walter Haidinger
https://bugzilla.novell.com/show_bug.cgi?id=335486
User kasievers@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c90
Kay Sievers
b) make it adjustable (/etc/init.d/boot mounts /dev using "mount -n -t tmpfs -o mode=0755 udev /dev" -- I'm going to test if the size is used when I add /dev with a different size to /etc/fstab)...
It should honor the fstab options by re-mounting the already mounted filesystem. We do that for some other filesystems too.
In case of a) -- which path we can use instead - /dev/.tmp? /var may be mounted on a separate disk so it can't be used too.
Better use some name private to your package, and delete the directory after it is no longer needed when the rootfs is available. We should not put new generic names like ".tmp" in /dev, and suggest people to share that. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c91
--- Comment #91 from Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c92
--- Comment #92 from Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User michael.d.taylor@exgate.tek.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c93
--- Comment #93 from Michael Taylor
https://bugzilla.novell.com/show_bug.cgi?id=335486
User peter-mailbox@web.de added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c94
--- Comment #94 from Peter Küppers
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c95
Marius Tomaschewski
Hello Marius, Hi!
on my server, I used the /etc/fstab line "tmpfs /dev/shm tmpfs size=1G 0 0" cause it was recommended for a SAP Testdrive (Netweaver 2004s with MaxDB 7.6).
There are various sapnotes with hints on tmpfs and SAP memory management for Linux systems. In the SAP Note 941735 (SAP memory management for 64-bit Linux systems), I found a solution to customize my system without the line in /etc/fstab. But there is another hint in this SAP Note: ... TMPFS With the STD implementation (SAP profile parameter es/implementation=std), the SAP Extended Memory is no longer stored in the TMPFS (under /dev/shm). However, the TMPFS is required by the Virtual Machine Container (VMC). For this reason, we still recommend the same configuration of the TMPFS: 75% (RAM + Swap) is still recommended as the size. ... So I understand, that in my case the "tmpfs /dev/shm tmpfs size=1G 0 0" is still relevant. If not, is this a question for Linux (so bugzilla.novell and SLES) or more for SAP? Question would be "How to configure the tmpfs for the VMC otherwise?" With the patch you recommended or "hard" size= in /etc/init.d/boot or...?
Since udev (something like 10.x), the complete /dev is a tmpfs. Before, /dev was (usually) a normal directory on root-fs with static device files and only the /dev/shm directory was a tmpfs. Because the complete /dev is a tmpfs (udev & init scripts are using very less of it [254K on my system]), it is not required any more to create a separate tmpfs for /dev/shm. But this does not mean, that it is never needed to adjust the default size (of 50% RAM). The patch in comment #91 allows to specify the size directly for /dev, adding a /etc/fstab line like: udev /dev tmpfs size=3g,mode=755 0 0 and to remove the /dev/shm mount entry. If setting it to "75% (RAM + Swap)" (as recommended above) makes sense or not, is completely another issue. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c96
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c97
--- Comment #97 from Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c98
--- Comment #98 from Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c99
--- Comment #99 from Marius Tomaschewski
Hi, Hi!
Because the problem reoccurs, I re-applied the 'fix' in /etc/init.d/network. But it did not help this time! I even restored fstab with the offending tmpfs line. Needless to say this did not help either.
Basically I have no clue left. The problem is back, and there is no cheap fix anymore. Am I the only one with this reoccuring problem?
Please take a look to the output of "cat /proc/mounts" if there is a separate /dev/shm mounted. Then take a look if there are "new-stamp-$ID" files in /dev/shm/sysconfig matching you network cards interface IDs in the "ip link show" output ("3: eth1:" => check if new-stamp-3 exists). -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c100
--- Comment #100 from Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c101
Marius Tomaschewski
Marius,
There is no /dev/shm mounted: [...]
OK.
In sysconfig new-stamp-2 and new-stamp-4 exists
root@Planhold:/home/marc> ls /dev/shm/sysconfig config-eth0 config-lo config-wlan0 new-stamp-2 new-stamp-4 ready-lo
Hmm... strange - it should work then. Please reinstall most recent sysconfig and udev RPMs for your distribution, verify the install using "rpm -V sysconfig udev" and reboot (true reboot). When it still happens, please enable the ". scripts/extradebug" line in the /sbin/ifup and /etc/init.d/network scripts and reboot (true reboot). Depending on the distribution/sysconfig version it should create "bash -x" trace files either in /tmp/exdeb.* or in /dev/shm/sysconfig/exdeb.*. Please tar them together with the /dev/shm/sysconfig files (tar cvzf bug335486-exdeb.tgz /tmp/exdeb.* /dev/shm/sysconfig) and attach to this bug. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User peter-mailbox@web.de added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c102
--- Comment #102 from Peter Küppers
Hi,
Because the problem reoccurs, I re-applied the 'fix' in /etc/init.d/network. But it did not help this time! I even restored fstab with the offending tmpfs line. Needless to say this did not help either.
Basically I have no clue left. The problem is back, and there is no cheap fix anymore. Am I the only one with this reoccuring problem?
Marc
Sorry, I'm late again with my answer. I updated the last upgrade packages on my server (openSUSE 10.3), but the solution (remove tmpfs from /etc/ftsab) still works on my server. cat /proc/mounts says, that there is no separate /dev/shm mounted. ip link show says there an no "new-stamp-$ID". in /dev/shm/sysconfig new-stamp-2 and new-stamp-3 exists (and both have 'cat' renamed). Cheers Peter -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c103
--- Comment #103 from Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User novell-web@zmi.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c104
--- Comment #104 from Michael Monnerie
fstab has a comment in front of tmpfs.
Marc, try completely deleting that entry. Even with a comment in front, it didn't work for me. Removing the line helped. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c105
--- Comment #105 from Marc Munnen
Marc, try completely deleting that entry. Even with a comment in front, it didn't work for me. Removing the line helped.
This looks like wizzardry. If that helps, it's magic. But it did not help for me, I am sorry to say. I will try to upgrade to OpenSuse 11.1, maybe that is more rewarding. Marc -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c106
--- Comment #106 from Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c107
--- Comment #107 from Marius Tomaschewski
Created an attachment (id=265808) --> (https://bugzilla.novell.com/attachment.cgi?id=265808) [details] Output from scripts/extradebug
Marius,
My first verify with rpm showed something hopeful: root@Planhold:/home/marc> rpm -V sysconfig udev S.5....T /etc/init.d/network S.5....T /etc/sysconfig/network/scripts/ifup-wireless This disappeared after reinstalling.
Unfortunately the bug did not... See the logfiles, Hopefully something evil will reveal itself.
Marc
The bug was not away, because you've /etc/sysconfig/network/ifcfg-eth6 and /etc/sysconfig/network/ifcfg-eth7 files in your system and then the network script is waiting for this not-existing hardware to appear. A "rm /etc/sysconfig/network/ifcfg-eth{6,7}" solves the problem without a need to update to 11.1. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=335486
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c108
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User mt@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c109
Marius Tomaschewski
https://bugzilla.novell.com/show_bug.cgi?id=335486
User m.van.munnen@hccnet.nl added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c110
--- Comment #110 from Marc Munnen
https://bugzilla.novell.com/show_bug.cgi?id=335486
User walter.haidinger@gmx.at added comment
https://bugzilla.novell.com/show_bug.cgi?id=335486#c111
--- Comment #111 from Walter Haidinger
participants (1)
-
bugzilla_noreply@novell.com