[Bug 1073433] Re: Ext4 corruption associated with shutdown of Ubuntu 12.10

[Bernd Schubert]

I fail to understand why so many people are insisting this is a kernel
bug. It isn't. What basically happens is that the network-manager
process is not killed by killall5. This process also has an open file
descriptor and therefore the root partitions cannot be mounted read-only
before shutdown, as the kernel reports it is busy. Network-manager will
not be killed due to the ubuntu patch add_sendsigs_omissions.patch,
which was added to NM for a similar reason:

From: Mathieu Trudel-Lapierre <mathieu.trudel-lapierre at canonical.com>
Subject: Move NM's spawns pid files to /run/sendsigs.omit.d.
Bug-Ubuntu: https://bugs.launchpad.net/ubuntu/+source/network-manager/+bug/869635

The pidfiles are only used by NM to start and be able to stop the same process,
their actual location doesn't overly matter. In this case, putting them
under /run/sendsigs.omit.d allows shutdown to skip over those as handled by
an upstart job, and in fact they are (although indirectly).

I have no idea what Mathieu actually intended with this patch, but it is entirely wrong and made everything worse. Instead of refusing to kill NM, it needs to be killed, which is just the other way around than what the patch is doing.
The only way to fix this from kernel point of view would be to write a kernel patch to allow to mount read-only while there are still file descriptors, which have write access. All those FDs would need to be closed from the kernel side and propably would cause application crashes.

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to upstart in Ubuntu.
https://bugs.launchpad.net/bugs/1073433

Title:
  Ext4 corruption associated with shutdown of Ubuntu 12.10

Status in Upstart:
  Confirmed
Status in “linux” package in Ubuntu:
  Incomplete
Status in “network-manager” package in Ubuntu:
  Triaged
Status in “upstart” package in Ubuntu:
  Confirmed

Bug description:
  1. Format and label a target Ext4 partion using Ubuntu 12.04
  2. Install 64bit 12.10 OS using that target without reformatting it
  3. Shut down 
  4. Boot an alternate copy of Ubuntu
  5. Restart selecting the newly installed OS
  6. Login then shutdown
  6. Boot an alternate copy of Ubuntu
  7.Fsck the newly installed OS allowing corrections to be made

  Each time the the newly installed OS is executed and then shutdown,
  even if execution only consists of logging on, a subsequent fsck will
  FAIL.

  I used Acronis True Image Home 2013 to create an image of the newly
  installed 64-bit Ubuntu 12.10, so I can recreate the symptoms of Ext4
  filesystem corruption 100% of the time by restoring from the image,
  booting, logging on and shutting down.

  ProblemType: Bug
  DistroRelease: Ubuntu 12.10
  Package: linux-image-3.5.0-17-generic 3.5.0-17.28
  ProcVersionSignature: Ubuntu 3.5.0-17.28-generic 3.5.5
  Uname: Linux 3.5.0-17-generic x86_64
  ApportVersion: 2.6.1-0ubuntu3
  Architecture: amd64
  AudioDevicesInUse:
   USER        PID ACCESS COMMAND
   /dev/snd/controlC1:  aguru      1871 F.... pulseaudio
   /dev/snd/controlC0:  aguru      1871 F.... pulseaudio
  CRDA: Error: command ['iw', 'reg', 'get'] failed with exit code 1: nl80211 not found.
  Date: Tue Oct 30 22:24:54 2012
  HibernationDevice: RESUME=UUID=f22e3fa5-c5c5-41f1-ae5a-49390547cb67
  InstallationMedia: Ubuntu 12.10 "Quantal Quetzal" - Release amd64 (20121017.5)
  IwConfig:
   eth0      no wireless extensions.
   
   lo        no wireless extensions.
  MachineType: System manufacturer P5Q-E
  ProcEnviron:
   PATH=(custom, no user)
   XDG_RUNTIME_DIR=<set>
   LANG=en_US.UTF-8
   SHELL=/bin/bash
  ProcFB: 0 nouveaufb
  ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-3.5.0-17-generic root=UUID=ef2c78d5-783a-422a-88f7-27ec09dda0d1 ro quiet splash vt.handoff=7
  RelatedPackageVersions:
   linux-restricted-modules-3.5.0-17-generic N/A
   linux-backports-modules-3.5.0-17-generic  N/A
   linux-firmware                            1.95
  RfKill:
   
  SourcePackage: linux
  UpgradeStatus: No upgrade log present (probably fresh install)
  dmi.bios.date: 04/06/2009
  dmi.bios.vendor: American Megatrends Inc.
  dmi.bios.version: 2101
  dmi.board.asset.tag: To Be Filled By O.E.M.
  dmi.board.name: P5Q-E
  dmi.board.vendor: ASUSTeK Computer INC.
  dmi.board.version: Rev 1.xx
  dmi.chassis.asset.tag: Asset-1234567890
  dmi.chassis.type: 3
  dmi.chassis.vendor: Chassis Manufacture
  dmi.chassis.version: Chassis Version
  dmi.modalias: dmi:bvnAmericanMegatrendsInc.:bvr2101:bd04/06/2009:svnSystemmanufacturer:pnP5Q-E:pvrSystemVersion:rvnASUSTeKComputerINC.:rnP5Q-E:rvrRev1.xx:cvnChassisManufacture:ct3:cvrChassisVersion:
  dmi.product.name: P5Q-E
  dmi.product.version: System Version
  dmi.sys.vendor: System manufacturer

To manage notifications about this bug go to:
https://bugs.launchpad.net/upstart/+bug/1073433/+subscriptions