On Tuesday I wanted to establish a VPN connection into the university network, but it didn’t work. A quick search on Launchpad revealed, that this was a bug in a recent upgrade to NetworkManager. The downstream maintainer said, one could either use the NetworkManager packages from the personal package archive of the downstream NetworkManager team or upgrade to Intrepid. I wanted to upgrade to Intrepid anyway, so why not now?
I backed up my home directory on an external hard disk and downloaded the Ubuntu alternative installer (64 bit version). Unfortunately, the first CD I had burned was corrupted. The second one was fine. Installation went ok. It didn’t find my network card, because of the e1000e issue, but as this was fixed in a recent Ubuntu kernel, I opted for a network-less installation.
Ater installation I downloaded the kernel packages for the recent 2.6.27.7 kernel, which has a fix for the e1000e issue, and transferred the files to my notebook via the external hard disk (Note to myself: Get a USB stick!). It was a bit strange to download packages for the architecture amd64 (as you can see from the e1000e thing, I’m on an Intel stack). Installation of the new kernel was successful and lsmod revealed, that the e1000e module was loaded.
Unfortunately, there still was no internet access. After some debugging, I found out that ping worked to all sites, but connections in Firefox or lynx only worked for the local network. Any attempt to connect to the outside net would just time out. By searching on Launchpad for ping works I found bug 11721. It turns out to be a slight incompatibility between the TCP/IP stack of my veteran Netgear router and that of the 2.6.27 kernel. Luckily, there is already a patch and there also is an easy workaround. Open a terminal and type in the following (you need root access to the machine):
sudo su -echo 0 > /proc/sys/net/ipv4/tcp_sackexit
It’s really sack, not stack as you might think.
Update: After further investigation upstream, there is now a better workaround for this issue. Please refer to the detailed description of the workaround at the Mandriva community wiki.
All that’s left to do for me now is transferring the rest of my home directory and installing some additional software packages.
Shameless plug: If this post was useful to you, please consider buying yourself something from one of my Amazon stores: US store, UK store, FR store, DE store, CA store. If you're not into Amazon, why not donate something to GNOME, Mozilla or Wikipedia? Thank you!
Why did you do a complete new installation instead of just upgrading from hardy?
I’m very curious about upgrading experiences as I’m planning that step on the next weekend, too.
By the way, I would have bought a new network card for long if I had to wait for a working module so long…
Actually, I did an upgrade first. It did work out okay, but I was hit by the tcp bug of course. As my system is kind of old (I had made numerous dist upgrades already) I thought this was an issue with the upgrade. Turned out it wasn’t.
This was not such a big problem, however. I wanted to do a fresh install anyway. I think one or two dist upgrades are generally ok, but from time to time you should do a fresh install, especially since it is not such a pain with Ubuntu as it’s e.g. with Windows. Also, I always have a number of third party packages installed, which makes upgrade not even easier. But (except for the tcp issue) the upgrade always worked for me (give in a few very small issues).
Additional bonus of the fresh install: New encryption method and passphrase.
Which algorithm? Not AES anymore? That would also be a reason for me to do a fresh install…
More information on upgrade issues can be found here:
http://www.ubuntu.com/getubuntu/releasenotes/810