Bug#765577: udev-udeb: Write dupliate entries into 70-persistent-net.rules
Hi all, Petter and Michael, thanks for your work on this! I'm able to reproduce this bug consistently (100% of the time, having run ~40 installations in the past week that exhibit it) with the following steps: 1. PXE boot a jessie netboot image (latest daily build, for example) 2. Add auto=true to the install menu option and run it. 3. Stop the installation (or switch to another TTY) and look at /etc/udev/rules.d/70-persistent-net.rules (in the installer environment, not under /target). The duplicate entry is present already, and will make its way into the target system during installation. The entry always appears for me when I use auto=true. I have not seen it appear without this option, but I haven't tested non-auto installs extensively. It happens even if we temporarily disable d-i preseeding from our DHCP server. On Tue, Oct 21, 2014 at 11:07:51AM +0200, Petter Reinholdtsen wrote: Perhaps it only happen on fairly quick machines? I experience it on a Thinkpad X200 and X300, but not on a 10 year old Latitude D505. :( It's possible that all of the hardware I'm testing on is too fast, but having performed many installs, I'm yet to see this not happen when auto=true is set. I've tested jessie installs on about 20 different systems (mix of ~15 new and old desktops, plus some KVM and VirtualBox guests), and seen this happen reliably on all of them. I wasn't able to narrow down the source of the duplicate entries in the code, despite some attempts at bisecting what I was hoping might be relevent parts of d-i. CCing debian-boot and adding the d-i tag, as I believe there's a good chance it is relevant to this bug. Apologies if that's unnecessary! Thanks and happy Monday, Chris -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#765577: udev-udeb: Write dupliate entries into 70-persistent-net.rules
[Petter Reinholdtsen]
It would be nice if /lib/udev/write_net_rules could be changed to log
a message to syslog when 70-persistent-net.rules is updated, to be
able to figure out when it happen. I in the places I suspect are
related to this issue, add similar instrumentation in the hw-detect
code.
Anyone around to help out with this issue?
While waiting for ideas how to find the source of the duplicate
entries, I added the following workaround to debian-edu-config,
rewriting the 70-persistent-net.rules file to remove duplicate if
anyone are present in the pre-pkgsel step. It solve the problem for
us for now, but I suspect a generic fix should be added to the code
inserting the duplicate entries.
diff --git a/share/debian-edu-config/d-i/pre-pkgsel
b/share/debian-edu-config/d-i/pre-pkgsel
index c298731..34d47a9 100644
--- a/share/debian-edu-config/d-i/pre-pkgsel
+++ b/share/debian-edu-config/d-i/pre-pkgsel
@@ -209,6 +209,13 @@ EOF
echo $MAILNAME /target/etc/mailname
fi
+# Workaround for bug in d-i/udev, causing duplicate entries in
+# /etc/udev/rules.d/70-persistent-net.rules in both d-i and
+# /target/. See bug $765577 for the details.
+if in-target /usr/share/debian-edu-config/tools/workaround-udev-bug-765577
; then
+ error disabled duplicate udev rule (bug #765577)
+fi
+
# Make sure that the interfaces are there for the cfengine run if
# network isn't already configured
if route | grep -q default ; then
diff --git a/share/debian-edu-config/tools/workaround-udev-bug-765577
b/share/debian-edu-config/tools/workar
new file mode 100755
index 000..7054c58
--- /dev/null
+++ b/share/debian-edu-config/tools/workaround-udev-bug-765577
@@ -0,0 +1,71 @@
+#!/usr/bin/perl
+#
+# Implement workaround for bug #765577 by removing duplicate entries
+# from the persistent network interface name rule before rebooting for
+# the first time, to make sure eth0 is present on the machines with
+# fixed network setup in /etc/network/interfaces
+
+use strict;
+use warnings;
+use Getopt::Std;
+
+my $bugurl = https://bugs.debian.org/765577;;
+
+my $rulefile = /etc/udev/rules.d/70-persistent-net.rules;
+my $newfile = $rulefile.new;
+
+my $debug = 0;
+my %linecache;
+my $modified = 0;
+my %opts;
+
+sub usage {
+my $retval = shift;
+print EOF;
+Usage: $0: [-dn]
+Fix
+ -d enable debugging
+ -n do not modify $rulefile
+EOF
+ exit($retval) if $retval;
+}
+
+getopts(dn, \%opts) || usage(1);
+
+open(my $rh, '', $rulefile) || die error: unable to read from $rulefile;
+my $wh;
+if (!$opts{'n'}) {
+open($wh, '', $newfile) || die error: unable to write to $newfile;
+}
+my $shortline;
+while (my $line = $rh) {
+$shortline = $line;
+$shortline =~ s/, NAME=[^]+//;
+print STDERR shortline: '$shortline'\n if $opts{'d'};
+if ($shortline !~ m/^\s*$/
+ $shortline =~ m/^SUBSYSTEM=/
+ exists $linecache{$shortline}) {
+# Seen the same line before, skip it.
+print STDERR skipping line\n if $opts{'d'};
+$modified = 1;
+if (!$opts{'n'}) {
+print $wh # Duplicate entry disabled, workaround for
$bugurl\n#$line;
+}
+next;
+}
+$linecache{$shortline} = 1;
+if (!$opts{'n'}) {
+print $wh $line;
+}
+}
+close($rh);
+if (!$opts{'n'}) {
+close($wh);
+if ($modified) {
+rename($newfile, $rulefile) || die error: unable to rename
$newfile to $rulefile;
+} else {
+unlink $newfile;
+}
+} else {
+}
+exit ! $modified;
--
Happy hacking
Petter Reinholdtsen
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#765577: udev-udeb: Write dupliate entries into 70-persistent-net.rules
[Petter Reinholdtsen] Could this be related to the fetching and installation of firmware in d-i? It normally cause kernel modules to be reloaded. I did some more testing, and when PXE booting the netboot d-i images, the 70-persistent-net.rules file got the duplicate entries very early. Only the localechooser, console-setup-udeb, ethdetect and netcfg menu items have been selected. The first is added before localechooser show its menu when I do not run the installation on automatic. But installing interactively do not cause the duplicate entries, and I am unable to insert instrumentation without doing things interactively. :( Perhaps it only happen on fairly quick machines? I experience it on a Thinkpad X200 and X300, but not on a 10 year old Latitude D505. :( It would be nice if /lib/udev/write_net_rules could be changed to log a message to syslog when 70-persistent-net.rules is updated, to be able to figure out when it happen. I in the places I suspect are related to this issue, add similar instrumentation in the hw-detect code. -- Happy hacking Petter Reinholdtsen -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#765577: udev-udeb: Write dupliate entries into 70-persistent-net.rules
[Michael Biebl] Could be a race. Maybe hwdetect triggers a uevent for your ethernet device, while /lib/udev/write_net_rules has not finished yet writing 70-persistent-net.rules and udev having picked up the changed 70-persistent-net.rules (via inotify). So instead of applying the generated rule for eth0, it generates a new rule, for eth1. Marco is the expert here though, so maybe he can shed some light on this. Great. I look forward to hearing from him. Could this be related to the fetching and installation of firmware in d-i? It normally cause kernel modules to be reloaded. -- Happy hacking Petter Reinholdtsen -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#765577: udev-udeb: Write dupliate entries into 70-persistent-net.rules
Package: udev-udeb
Version: 215-5+b1
Severity: important
User: debian-...@lists.debian.org
Usertags: debian-edu
I ran into this problem when installing a Debian Edu Jessie workstation
using PXE on a Lenovo Thinkpad X200. The problem at the end is that
users can not log into the workstation because the network interface is
not properly configured. This is caused by the network card to show up
as eth1 instead of eth0. Only eth0 is configured in
/etc/network/interfaces.
This in turn is caused by /etc/udev/rules.d/70-presisten-net.rules
listing the network card twice, once with eth0 and once with eth1. The
file look like this:
# This file was automatically generated by the /lib/udev/write_net_rules
# program, run by the persistent-net-generator.rules rules file.
#
# You can modify it, as long as you keep each rule on a single
# line, and change only the value of the NAME= key.
# PCI device 0x8086:0x10f5 (e1000e)
SUBSYSTEM==net, ACTION==add, DRIVERS==?*,
ATTR{address}==00:1f:16:0b:47:4d, ATTR{dev_id}==0x0, ATTR{type}==1,
KERNEL==eth*, NAME=eth0
# PCI device 0x8086:0x10f5 (e1000e)
SUBSYSTEM==net, ACTION==add, DRIVERS==?*,
ATTR{address}==00:1f:16:0b:47:4d, ATTR{dev_id}==0x0, ATTR{type}==1,
KERNEL==eth*, NAME=eth1
As this have not happend to me every time, I have tried to figure out
what is special about this setup, but do not know the udev stuff enough
to have a clue. I do notice this part in the log, which seem to be
related:
Oct 16 10:51:35 main-menu[186]: INFO: Menu item 'ethdetect' selected
Oct 16 10:51:35 kernel: [2.532038] usb 2-2: new full-speed USB device
number 3 using uhci_hcd
Oct 16 10:51:35 net/hw-detect.hotplug: Detected hotpluggable network interface
lo
Oct 16 10:51:35 net/hw-detect.hotplug: Detected hotpluggable network interface
eth1
Oct 16 10:51:35 kernel: [2.620227] systemd-udevd[51]: renamed network
interface eth0 to eth1
Oct 16 10:51:35 kernel: [2.636070] Switched to clocksource tsc
Oct 16 10:51:35 kernel: [2.707037] usb 2-2: New USB device found,
idVendor=0a5c, idProduct=2145
Oct 16 10:51:35 kernel: [2.707039] usb 2-2: New USB device strings: Mfr=1,
Product=2, SerialNumber=0
Oct 16 10:51:35 kernel: [2.707042] usb 2-2: Product: ThinkPad Bluetooth
with Enhanced Data Rate II
Oct 16 10:51:35 kernel: [2.707043] usb 2-2: Manufacturer: Lenovo Computer
Corp
Oct 16 10:51:35 hw-detect: Detected module 'usb-storage' for 'USB storage'
Notice how eth0 was renamed to eth1 by udev while hw-detect version
1.104 was running. Any clue what is going wrong here?
--
Happy hacking
Petter Reinholdtsen
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#765577: udev-udeb: Write dupliate entries into 70-persistent-net.rules
Am 16.10.2014 um 13:09 schrieb Petter Reinholdtsen:
Package: udev-udeb
Version: 215-5+b1
Severity: important
User: debian-...@lists.debian.org
Usertags: debian-edu
I ran into this problem when installing a Debian Edu Jessie workstation
using PXE on a Lenovo Thinkpad X200. The problem at the end is that
users can not log into the workstation because the network interface is
not properly configured. This is caused by the network card to show up
as eth1 instead of eth0. Only eth0 is configured in
/etc/network/interfaces.
This in turn is caused by /etc/udev/rules.d/70-presisten-net.rules
listing the network card twice, once with eth0 and once with eth1. The
file look like this:
# This file was automatically generated by the /lib/udev/write_net_rules
# program, run by the persistent-net-generator.rules rules file.
#
# You can modify it, as long as you keep each rule on a single
# line, and change only the value of the NAME= key.
# PCI device 0x8086:0x10f5 (e1000e)
SUBSYSTEM==net, ACTION==add, DRIVERS==?*,
ATTR{address}==00:1f:16:0b:47:4d, ATTR{dev_id}==0x0, ATTR{type}==1,
KERNEL==eth*, NAME=eth0
# PCI device 0x8086:0x10f5 (e1000e)
SUBSYSTEM==net, ACTION==add, DRIVERS==?*,
ATTR{address}==00:1f:16:0b:47:4d, ATTR{dev_id}==0x0, ATTR{type}==1,
KERNEL==eth*, NAME=eth1
As this have not happend to me every time, I have tried to figure out
what is special about this setup, but do not know the udev stuff enough
to have a clue. I do notice this part in the log, which seem to be
related:
Oct 16 10:51:35 main-menu[186]: INFO: Menu item 'ethdetect' selected
Oct 16 10:51:35 kernel: [2.532038] usb 2-2: new full-speed USB device
number 3 using uhci_hcd
Oct 16 10:51:35 net/hw-detect.hotplug: Detected hotpluggable network
interface lo
Oct 16 10:51:35 net/hw-detect.hotplug: Detected hotpluggable network
interface eth1
Oct 16 10:51:35 kernel: [2.620227] systemd-udevd[51]: renamed network
interface eth0 to eth1
Oct 16 10:51:35 kernel: [2.636070] Switched to clocksource tsc
Oct 16 10:51:35 kernel: [2.707037] usb 2-2: New USB device found,
idVendor=0a5c, idProduct=2145
Oct 16 10:51:35 kernel: [2.707039] usb 2-2: New USB device strings:
Mfr=1, Product=2, SerialNumber=0
Oct 16 10:51:35 kernel: [2.707042] usb 2-2: Product: ThinkPad Bluetooth
with Enhanced Data Rate II
Oct 16 10:51:35 kernel: [2.707043] usb 2-2: Manufacturer: Lenovo Computer
Corp
Oct 16 10:51:35 hw-detect: Detected module 'usb-storage' for 'USB storage'
Notice how eth0 was renamed to eth1 by udev while hw-detect version
1.104 was running. Any clue what is going wrong here?
Could be a race. Maybe hwdetect triggers a uevent for your ethernet
device, while /lib/udev/write_net_rules has not finished yet writing
70-persistent-net.rules and udev having picked up the changed
70-persistent-net.rules (via inotify). So instead of applying the
generated rule for eth0, it generates a new rule, for eth1.
Marco is the expert here though, so maybe he can shed some light on this.
--
Why is it that all of the instruments seeking intelligent life in the
universe are pointed away from Earth?
signature.asc
Description: OpenPGP digital signature
