2

I am experiencing kernel panic when trying to mount a ext4 filesystem which itself resides on a truecrypt volume mounted using -m nokernelcrypto option.

The system is CentOS 6.3, OpenVZ Kernel 2.6.32-042stab074.10 The disk is a virtual disk (a truecrypt volume mounted with --filesystem=none), and the actual command triggering kernel panic is:

mount -o loop -t ext4 /tmp/.truecrypt_aux_mnt2/volume /mnt/storage

Note that:

  1. It mounts fine in read-only mode, i. e. using -m ro
  2. It also mounts perfectly fine if I boot in "normal" kernel (non-OpenVZ), which is 2.6.32-279.22.1.el6.x86_64
  3. Again, it mounts perfectly fine when not using -m nokernelcrypto (either kernel)
  4. I have to use -m nokernelcrypto because a 3TB disk that I have can only be mounted using that option — otherwise truecrypt refuses to mount saying "Kernel cryptographic service test failed" and suggests to upgrade kernel to 2.6.33. Not using -m nokernelcrypto goes OK with 2TB disks, but not with 3TB
  5. The panic message features "Kernel panic - not syncing: Fatal exception" and "kernel BUG at fs/fuse/file.c:1110!"
  6. I did not have this problem when I was using Debian OpenVZ — was able to mount the 3TB using -m nokernelcrypto easily

I am just not sure where exactly the problem is: CentOS, OpenVZ, fuse library or the non-kernel decryption. Any clues will be much appreciated.

Anton
  • 21
  • 1

1 Answers1

1

I have a similar problem using Proxmox (see pve.proxmox.com, it's a Debian derivative). I have an ext4 file system disk image residing on a glusterfs share (which is fuse-based). As soon as I mount the disk image, I get the message about the kernel bug at fs/fuse/file.c:1110.

This all worked fine with pve-kernel-2.6.32-11-pve. I ran into the problem after upgrading the kernel to pve-kernel-2.6.32-19-pve.

So I guess the problem might occur whenever we mount a disk images that is hosted on a fuse-based file system on a recent (whatever that means) 2.6.32 kernel.

Johannes Martin
  • 11
  • 1