Kubuntu 12.10 bcmwl install failure

This bug has been reported on the Ubuntu ISO testing tracker.

A list of all reports related to this bug can be found here:
http://iso.qa.ubuntu.com/qatracker/reports/bugs/1065827

It's in the pool in the ISO:

/pool/restricted/b/bcmwl/bcmwl-kernel-source_5.100.82.112+bdcom-0ubuntu2_i386.deb

There is no CD source in /etc/apt/sources.list, so it's not looking in the local pool.

It didn't find the CD (this is a USB, actually) repository to install non-free drivers during install either.

Turns out this is a bug in building the images.

This is actually a casper regression from bug 909919, and one we need to fix for release.

There is something else going on here which I don't understand. After fixing the casper regression, software-properties -> Additional Drivers is able to install packages from the CD and thus correctly installs bcmwl-kernel-source on demand. However, it fails to actually load the wl module; and, if I load that module by hand, I still don't appear to get a new network interface. There seem to be multiple problems here.

This happens with both u-d-common as well as jockey, but both do nothing more than just apt-get installing the package. Apparently there's something wrong with the module? I'm testing this now on my Atom netbook which has a Broadcom wifi card.

This bug was fixed in the package casper - 1.328

---------------
casper (1.328) quantal; urgency=low

  * Fix path to .disk/info in 41apt_cdrom (LP: #1065827).
 -- Colin Watson <email address hidden> Fri, 12 Oct 2012 12:05:15 +0100

So apparently we have two bugs in bcmwl:

- postinst needs to "modprobe wl", or more elegantly, call udevadm trigger on broadcom wlan devices to get the module loaded.
- even after "modprobe wl" the driver doesn't detect any devices; that definitively worked in earlier releases on my netbook (14E4:4315), so this might be a kernel/driver version incompatiblity?

The second point is because the b43 driver claims the device by default. Right now the package is designed to work after a reboot, but if we want it to work right after installation, we'd need something like

  rmmod b43 ssb || :
  modprobe wl || :

at the end of the postinst.

Uploaded fixed bcmwl to quantal unapproved.

This bug was fixed in the package bcmwl - 5.100.82.112+bdcom-0ubuntu3

---------------
bcmwl (5.100.82.112+bdcom-0ubuntu3) quantal; urgency=low

  * debian/bcmwl-kernel-source.postinst: Mirror the actions of the installed
    modprobe.d files, i. e. rmmod the b43 driver and try to modprobe wl, so
    that this package works right after installation. (LP: #1065827)
  * debian/bcmwl-kernel-source.postinst: Drop duplicate #DEBHELPER# token. It
    does not currently expand to anything, but it looks confusing.
 -- Martin Pitt <email address hidden> Fri, 12 Oct 2012 14:00:38 +0200

Hello Scott, or anyone else affected,

Accepted bcmwl into precise-proposed. The package will build now and be available at http://launchpad.net/ubuntu/+source/bcmwl/6.20.155.1+bdcom-0ubuntu0.0.1 in a few hours, and then in the -proposed repository.

Please help us by testing this new package. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Your feedback will aid us getting this update out to other Ubuntu users.

If this package fixes the bug for you, please add a comment to this bug, mentioning the version of the package you tested, and change the tag from verification-needed to verification-done. If it does not fix the bug for you, please add a comment stating that, and change the tag to verification-failed. In either case, details of your testing will help us make a better decision.

Further information regarding the verification process can be found at https://wiki.ubuntu.com/QATeam/PerformingSRUVerification . Thank you in advance!

Since the bug is for a no network install, how do I test this?

I'd think something along this should work:
1. Put http://cdimage.ubuntu.com/precise/daily-live/current/ (or http://cdimage.ubuntu.com/kubuntu/precise/daily-live/current/) i386 version onto bootable USB stick
2. Replace the bcml-kernel-source .deb at pool/restricted/b/bcmwl/ with a .deb from https://launchpad.net/ubuntu/+source/bcmwl/6.20.155.1+bdcom-0ubuntu0.0.1/+build/4226259
3. Boot.
4. I guess the installation needs to be mimicked with dpkg -i, but then again the actual installation/image bug supposedly was only in quantal and already fixed, and the remaining bug that needs now verification in precise is solely about the driver package itself.

Hi!

I did a test of this in a Dell Vostro 1520, which has this broadcom chip:

0e:00.0 Network controller: Broadcom Corporation BCM4322 802.11a/b/g/n Wireless LAN Controller (rev 01)

This can use either b43 or wl. First, I did a 12.04.1 installation, with no updates, and no 3rd-party driver installation.

Then I manually installed firmware-b43-installer. This pulled a bunch of firmware files and after that I was able to connect to a wireless network, the module loaded OK.

With things working like this, I installed bcmwl-kernel-source, it pulled version 5.100.82.38+bdcom-0ubuntu6.1 from precise-updates. It compiled the kernel module, the network interface disconnected for a moment and then came back up. I verified the wl module was loaded. I'm still able to access the network.

OK, next I removed (purge) both bcmwl-kernel-source and firmware-b43-installer. The b43 module tries to load but complains about missing firmware. I reinstalled firmware-b43-installer and went back to the point where b43 is working.

I added the -proposed repository and reinstalled bcmwl-kernel-source. It downloaded and installed version 6.20.155.1+bdcom-0ubuntu0.0.1. Again, the interface went down for a moment, then automatically back up with the new module. I confirmed with lsmod that wl is loaded and b43 is not. I'm still able to ping and access the network.

So even though I was unable to reproduce the original failure, at least the new package seems to behave correctly in this scenario, this should cover against the most severe regressions.

I worked with Daniel on the testing in comment #18. Based on the fact that the driver loads automatically upon installation (as described in the changelog entry related to this bug) I'm marking this as verification-done

The verification of this Stable Release Update has completed successfully and the package has now been released to -updates. Subsequently, the Ubuntu Stable Release Updates Team is being unsubscribed and will not receive messages about this bug report. In the event that you encounter a regression using the package from -updates please report a new bug using ubuntu-bug and tag the bug report regression-update so we can easily find any regresssions.

This bug was fixed in the package bcmwl - 6.20.155.1+bdcom-0ubuntu0.0.1

---------------
bcmwl (6.20.155.1+bdcom-0ubuntu0.0.1) precise-proposed; urgency=low

  [ Alberto Milone ]
  * New upstream release (LP: #923809).
  * Add 0004-Add-support-for-Linux-3.2.patch:
    - Make sure the patch can always be applied.
  * Add 0005-Add-support-for-Linux-3.4.0.patch:
    - Thanks to Eugene San for the patch (LP: #994255).
  * debian/control:
    - Drop all the dependencies on the linux-headers.
  * debian/dkms.conf.in:
    - Always apply all patches (LP: #1020059).

  [ Martin Pitt ]
  * debian/bcmwl-kernel-source.postinst:
    - Silence the errors from rmmod about the modules not being
      loaded. This is fully expected, and just causes confusion,
      as well as autopkgtest failures.
    - Mirror the actions of the installed modprobe.d files,
      i.e. rmmod the b43 driver and try to modprobe wl, so
      that this package works right after installation.
      (LP: #1065827)
    - Drop duplicate #DEBHELPER# token. It does not currently
      expand to anything, but it looks confusing.
  * debian/rules, debian/bcmwl-kernel-source.modaliases:
    - Drop parsing of src/src/wl/sys/wl_linux.c, as the current
      upstream version dropped all useful vendor matches in its
      pci_device_id struct and now has a broken modalias. Keep
      a static debian/bcmwl-kernel-source.modaliases instead
      which matches on any Broadcom network device (vendor 14E4
      class 02 subclass 80). This brings back a proper
      "Modaliases:" package header (LP: #1089943).
 -- Alberto Milone <email address hidden> Thu, 17 Jan 2013 12:38:17 +0100