kernel-packages team mailing list archive

[Dexuan Cui <decui@xxxxxxxxxxxxx>]

I believe it’s a configuration issue:

If we use the default Ubuntu kernel as the ”dump capture kernel“[1], we
need to reserve more memory – the default 128MB, generated by the linux-
crashdump post-install script, is not enough, causing a -ENOMEM
failure/BUG in a function of tree_plugin.h.

This issue even exists in the native case for Ubuntu 14.10. 
And essentially, Ubuntu 14.04 has the same issue, but probably the kernel/initramfs of 14.04 aren’t so big that the early kernel panic at “kernel BUG at /build/buildd/linux-3.16.0/kernel/rcu/tree_plugin.h:2332” doesn’t happen – however, at a later stage of 14.04 boot-up, many Out-of-Memory warnings are printed in my tests, though the crashdump can be created.

If I change the size of the default memory reserved for kdump from 128MB to 256MB, the issue won’t appear.
How to change  the size[2] – please update the grub.conf and change the kernel parameter [3] crashkernel=384M-:128M to crashkernel=384M-:256M.
 
References:
[1] https://wiki.archlinux.org/index.php/Kdump
[2] http://docs.oracle.com/cd/E19140-01/html/821-0281/gkvnc.html
[3] The syntax of the kernel parameter crashkernel: https://www.kernel.org/doc/Documentation/kernel-parameters.txt

-- 
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/1359980

Title:
  [Hyper-V] Unable to perform a full kernel crash on Ubuntu 14.10

Status in “linux” package in Ubuntu:
  In Progress
Status in “linux” source package in Utopic:
  In Progress

Bug description:
  System details:
  Linux ubuntu 3.16.0-9-generic #14-Ubuntu SMP Fri Aug 15 15:03:57 UTC 2014 x86_64 x86_64 x86_64 GNU/Linux
  Ubuntu Server 14.10 Daily build from 19th Aug (first build picked up for testing)

  We have followed the steps at
  https://wiki.ubuntu.com/Kernel/CrashdumpRecipe to configure kdump on a
  Generation 1 VM.

  kdump is loaded and when we trigger the crash, a kernel panic is
  displayed.

  Issues:
  1. The vmcore/crash file is not generated under /var/crash as configured in the kdump config.
  2. The VM will hang and not reboot.  Default action to take in case of a failure is to reboot - again as set in kdump config.

  Repro rate: 100%
  VM settings: tests ran with 1 CPU core or 4 cores.
  Negociated VMBus is 3.0, which has been integrated since 14.04 so this should work as no other patches are needed. (https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1317697)

  I'm also attaching the serial console output of both cases - single
  core and SMP.

  root@ubuntu:~# kdump-config  load
  efi memory descriptor version 0 is not supported!
   * loaded kdump kernel

  root@ubuntu:~# cat /sys/kernel/kexec_crash_loaded
  1

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1359980/+subscriptions