2.6.27 REGRESSION, hangs during boot while preparing restricted drivers when ath_hal is blacklisted

Bug #262539 reported by Matt Zimmerman
18
Affects Status Importance Assigned to Milestone
linux-restricted-modules (Ubuntu)
Fix Released
High
Ben Collins
Intrepid
Fix Released
High
Ben Collins

Bug Description

I was running happily with 2.6.26, upgraded to 2.6.27-1.2, and my system failed to boot. It hung during /etc/init.d/linux-restricted-modules-common, and this was reproducible after a hard reset. Booting 2.6.26 via "last successful boot" works fine.

I will attach screenshots showing a sysrq+t trace at the point where it hangs.

ProblemType: Bug
Architecture: i386
DistroRelease: Ubuntu 8.10
NonfreeKernelModules: nvidia
Package: linux-image-2.6.26-5-generic 2.6.26-5.17 [modified: lib/modules/2.6.26-5-generic/modules.pcimap lib/modules/2.6.26-5-generic/modules.dep lib/modules/2.6.26-5-generic/modules.ieee1394map lib/modules/2.6.26-5-generic/modules.usbmap lib/modules/2.6.26-5-generic/modules.isapnpmap lib/modules/2.6.26-5-generic/modules.seriomap lib/modules/2.6.26-5-generic/modules.alias lib/modules/2.6.26-5-generic/modules.symbols]
ProcCmdLine: root=UUID=daf1bf61-d974-4e90-b2c7-5f4dfa4e56cf ro quiet splash crashkernel=384M-2G:64M@16M,2G-:128M@16M last-good-boot
ProcEnviron:
 LC_COLLATE=C
 PATH=/home/username/bin:/usr/lib/ccache:/usr/local/bin:/usr/bin:/bin:/usr/bin/X11:/sbin:/usr/sbin:/usr/games
 LANG=en_US.UTF-8
 SHELL=/bin/zsh
ProcVersionSignature: Ubuntu 2.6.26-5.15-generic
SourcePackage: linux

Revision history for this message
Matt Zimmerman (mdz) wrote :
Matt Zimmerman (mdz)
Changed in linux:
importance: Undecided → High
Revision history for this message
Matt Zimmerman (mdz) wrote :

This looks like it's actually a userland issue in lrm-manager. The system comes up OK if I disable /etc/init.d/linux-restricted-modules-common, and I can see it go into an infinite loop when I run it manually thereafter.

The fact that lrm-manager spins this way does block the entire boot, though, so this is still a serious issue. I don't know what it is about 2.6.27 which triggers this. I'm attaching a sh -x trace showing the loop; it continues indefinitely in a similar pattern and never exits.

Revision history for this message
Matt Zimmerman (mdz) wrote :

The problem is with the DISABLED_MODULES checking. I use:

DISABLED_MODULES="ath_hal fc fglrx ltm"

(which has worked fine for ages). The problem is that every time it gets to "ath_hal" in the list, it adds "ath_hal" back:

    case "$1" in
      madwifi|ath_hal)
        set -- $@ ath_hal ath_pci ath_rate_amrr ath_rate_minstrel \
        ath_rate_onoe ath_rate_sample wlan wlan_acl \
        wlan_ccmp wlan_scan_ap wlan_scan_sta wlan_tkip \
        wlan_wep wlan_xauth
        shift

resulting in an infinite loop. I have no idea why this doesn't fail under 2.6.26. Running sudo lrm-manager --kver=2.6.26-5-generic --quick certainly does fail (running under 2.6.27).

Revision history for this message
John Dong (jdong) wrote :

Indeed this appears the case; I too am experiencing this when I have ath_hal blacklisted.

I've tried:

      madwifi)
        set -- $@ ath_hal
 shift
 ;;
      ath_hal)
       set -- $@ ath_pci ath_rate_amrr ath_rate_minstrel \
  ath_rate_onoe ath_rate_sample wlan wlan_acl \
  wlan_ccmp wlan_scan_ap wlan_scan_sta wlan_tkip \
  wlan_wep wlan_xauth
        shift
        ;;

Which works, but I'm not confident is the correct fix.

Changed in linux-restricted-modules:
assignee: nobody → ben-collins
milestone: none → intrepid-alpha-5
status: New → Fix Committed
Revision history for this message
Sindhudweep Sarkar (sindhudweep-sarkar) wrote :

Had same issue after upgrading to intrepid while using a custom ath_hal for my ATH5006x wifi card.

Revision history for this message
Grzegorz Sterniczuk (dzikus) wrote :

Me too, but with DISABLED_MODULES="madwifi fc ltm nv wl".

Revision history for this message
Steve Langasek (vorlon) wrote :

Postponed to alpha-6; this issue is specific to upgrades from Ubuntu 8.04 in which the user has blacklisted the module, and should not block the alpha milestone.

Changed in linux-restricted-modules:
milestone: intrepid-alpha-5 → intrepid-alpha-6
Revision history for this message
Launchpad Janitor (janitor) wrote :

This bug was fixed in the package linux-restricted-modules - 2.6.27-2.3

---------------
linux-restricted-modules (2.6.27-2.3) intrepid; urgency=low

  [Ben Collins]

  * build/lrm-manager: Fix endless loop for ath_hal
    - LP: #262539
  * broadcom: Force vlan_mode to off on initialization of device
  * build/control: Provides/Conflicts/Replaces atmel-firmware
    - LP: #263021

 -- Ben Collins <email address hidden> Fri, 05 Sep 2008 08:56:26 -0400

Changed in linux-restricted-modules:
status: Fix Committed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Duplicates of this bug

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.