Kernel Crash Dump

Introduction

A Kernel Crash Dump refers to a portion of the contents of volatile memory (RAM) that is copied to disk whenever the execution of the kernel is disrupted. The following events can cause a kernel disruption :

  • Kernel Panic

  • Non Maskable Interrupts (NMI)

  • Machine Check Exceptions (MCE)

  • Hardware failure

  • Manual intervention

For some of those events (panic, NMI) the kernel will react automatically and trigger the crash dump mechanism through kexec. In other situations a manual intervention is required in order to capture the memory. Whenever one of the above events occurs, it is important to find out the root cause in order to prevent it from happening again. The cause can be determined by inspecting the copied memory contents.

Kernel Crash Dump Mechanism

When a kernel panic occurs, the kernel relies on the kexec mechanism to quickly reboot a new instance of the kernel in a pre-reserved section of memory that had been allocated when the system booted (see below). This permits the existing memory area to remain untouched in order to safely copy its contents to storage.

Installation

The kernel crash dump utility is installed with the following command:

sudo apt install linux-crashdump

Starting with 16.04, the kernel crash dump mechanism is enabled by default. During the installation, you will be prompted with the following dialog. Unless chosen otherwise, the kdump mechanism will be enabled.

 |------------------------| Configuring kdump-tools |------------------------|
 |                                                                           |
 |                                                                           |
 | If you choose this option, the kdump-tools mechanism will be enabled. A   |
 | reboot is still required in order to enable the crashkernel kernel        |
 | parameter.                                                                |
 |                                                                           |
 | Should kdump-tools be enabled by default?                                 |
 |                                                                           |
 |                    <Yes>                       <No>                       |
 |                                                                           |
 |---------------------------------------------------------------------------|
        

If you ever need to manually enable the functionality, you can use the dpkg-reconfigure kdump-tools command and answer Yes to the question. You can also edit /etc/default/kdump-tools by including the following line:

USE_KDUMP=1

If a reboot has not been done since installation of the linux-crashdump package, a reboot will be required in order to activate the crashkernel= boot parameter. Upon reboot, kdump-tools will be enabled and active.

If you enable kdump-tools after a reboot, you will only need to issue the kdump-config load command to activate the kdump mechanism.

Configuration

In addition to local dump, it is now possible to use the remote dump functionality to send the kernel crash dump to a remote server, using either the SSH or NFS protocols.

Local Kernel Crash Dumps

Local dumps are configured automatically and will remain in use unless a remote protocol is chosen. Many configuration options exist and are thoroughly documented in the /etc/default/kdump-tools file.

Remote Kernel Crash Dumps using the SSH protocol

To enable remote dumps using the SSH protocol, the /etc/default/kdump-tools must be modified in the following manner :

# ---------------------------------------------------------------------------
# Remote dump facilities:
# SSH - username and hostname of the remote server that will receive the dump
#       and dmesg files.
# SSH_KEY - Full path of the ssh private key to be used to login to the remote
#           server. use kdump-config propagate to send the public key to the
#           remote server
# HOSTTAG - Select if hostname of IP address will be used as a prefix to the
#           timestamped directory when sending files to the remote server.
#           'ip' is the default.
SSH="ubuntu@kdump-netcrash"
        

The only mandatory variable to define is SSH. It must contain the username and hostname of the remote server using the format {username}@{remote server}.

SSH_KEY may be used to provide an existing private key to be used. Otherwise, the kdump-config propagate command will create a new keypair. The HOSTTAG variable may be used to use the hostname of the system as a prefix to the remote directory to be created instead of the IP address.

The following example shows how kdump-config propagate is used to create and propagate a new keypair to the remote server :

sudo kdump-config propagate
Need to generate a new ssh key...
The authenticity of host 'kdump-netcrash (192.168.1.74)' can't be established.
ECDSA key fingerprint is SHA256:iMp+5Y28qhbd+tevFCWrEXykDd4dI3yN4OVlu3CBBQ4.
Are you sure you want to continue connecting (yes/no)? yes
ubuntu@kdump-netcrash's password: 
propagated ssh key /root/.ssh/kdump_id_rsa to server ubuntu@kdump-netcrash
        
The password of the account used on the remote server will be required in order to successfully send the public key to the server

The kdump-config show command can be used to confirm that kdump is correctly configured to use the SSH protocol :

kdump-config show
DUMP_MODE:        kdump
USE_KDUMP:        1
KDUMP_SYSCTL:     kernel.panic_on_oops=1
KDUMP_COREDIR:    /var/crash
crashkernel addr: 0x2c000000
   /var/lib/kdump/vmlinuz: symbolic link to /boot/vmlinuz-4.4.0-10-generic
kdump initrd: 
   /var/lib/kdump/initrd.img: symbolic link to /var/lib/kdump/initrd.img-4.4.0-10-generic
SSH:              ubuntu@kdump-netcrash
SSH_KEY:          /root/.ssh/kdump_id_rsa
HOSTTAG:          ip
current state:    ready to kdump
        

Remote Kernel Crash Dumps using the NFS protocol

To enable remote dumps using the NFS protocol, the /etc/default/kdump-tools must be modified in the following manner :

# NFS -     Hostname and mount point of the NFS server configured to receive
#           the crash dump. The syntax must be {HOSTNAME}:{MOUNTPOINT} 
#           (e.g. remote:/var/crash)
#
NFS="kdump-netcrash:/var/crash"
          

As with the SSH protocol, the HOSTTAG variable can be used to replace the IP address by the hostname as the prefix of the remote directory.

The kdump-config show command can be used to confirm that kdump is correctly configured to use the NFS protocol :

kdump-config show
DUMP_MODE:        kdump
USE_KDUMP:        1
KDUMP_SYSCTL:     kernel.panic_on_oops=1
KDUMP_COREDIR:    /var/crash
crashkernel addr: 0x2c000000
   /var/lib/kdump/vmlinuz: symbolic link to /boot/vmlinuz-4.4.0-10-generic
kdump initrd: 
   /var/lib/kdump/initrd.img: symbolic link to /var/lib/kdump/initrd.img-4.4.0-10-generic
NFS:              kdump-netcrash:/var/crash
HOSTTAG:          hostname
current state:    ready to kdump
      

Verification

To confirm that the kernel dump mechanism is enabled, there are a few things to verify. First, confirm that the crashkernel boot parameter is present (note: The following line has been split into two to fit the format of this document:

cat /proc/cmdline

BOOT_IMAGE=/vmlinuz-3.2.0-17-server root=/dev/mapper/PreciseS-root ro
 crashkernel=384M-2G:64M,2G-:128M

The crashkernel parameter has the following syntax:

crashkernel=<range1>:<size1>[,<range2>:<size2>,...][@offset]
    range=start-[end] 'start' is inclusive and 'end' is exclusive.
        

So for the crashkernel parameter found in /proc/cmdline we would have :

crashkernel=384M-2G:64M,2G-:128M

The above value means:

  • if the RAM is smaller than 384M, then don't reserve anything (this is the "rescue" case)

  • if the RAM size is between 386M and 2G (exclusive), then reserve 64M

  • if the RAM size is larger than 2G, then reserve 128M

Second, verify that the kernel has reserved the requested memory area for the kdump kernel by doing:

dmesg | grep -i crash

...
[    0.000000] Reserving 64MB of memory at 800MB for crashkernel (System RAM: 1023MB)

Finally, as seen previously, the kdump-config show command displays the current status of the kdump-tools configuration :

        kdump-config show
DUMP_MODE:        kdump
USE_KDUMP:        1
KDUMP_SYSCTL:     kernel.panic_on_oops=1
KDUMP_COREDIR:    /var/crash
crashkernel addr: 0x2c000000
   /var/lib/kdump/vmlinuz: symbolic link to /boot/vmlinuz-4.4.0-10-generic
kdump initrd: 
      /var/lib/kdump/initrd.img: symbolic link to /var/lib/kdump/initrd.img-4.4.0-10-generic
current state:    ready to kdump

kexec command:
      /sbin/kexec -p --command-line="BOOT_IMAGE=/vmlinuz-4.4.0-10-generic root=/dev/mapper/VividS--vg-root ro debug break=init console=ttyS0,115200 irqpoll maxcpus=1 nousb systemd.unit=kdump-tools.service" --initrd=/var/lib/kdump/initrd.img /var/lib/kdump/vmlinuz
      

Testing the Crash Dump Mechanism

Testing the Crash Dump Mechanism will cause a system reboot. In certain situations, this can cause data loss if the system is under heavy load. If you want to test the mechanism, make sure that the system is idle or under very light load.

Verify that the SysRQ mechanism is enabled by looking at the value of the /proc/sys/kernel/sysrq kernel parameter :

cat /proc/sys/kernel/sysrq

If a value of 0 is returned the feature is disabled. Enable it with the following command :

sudo sysctl -w kernel.sysrq=1

Once this is done, you must become root, as just using sudo will not be sufficient. As the root user, you will have to issue the command echo c > /proc/sysrq-trigger. If you are using a network connection, you will lose contact with the system. This is why it is better to do the test while being connected to the system console. This has the advantage of making the kernel dump process visible.

A typical test output should look like the following :

sudo -s
[sudo] password for ubuntu: 
# echo c > /proc/sysrq-trigger
[   31.659002] SysRq : Trigger a crash
[   31.659749] BUG: unable to handle kernel NULL pointer dereference at           (null)
[   31.662668] IP: [<ffffffff8139f166>] sysrq_handle_crash+0x16/0x20
[   31.662668] PGD 3bfb9067 PUD 368a7067 PMD 0 
[   31.662668] Oops: 0002 [#1] SMP 
[   31.662668] CPU 1 
....

The rest of the output is truncated, but you should see the system rebooting and somewhere in the log, you will see the following line :

Begin: Saving vmcore from kernel crash ...
Once completed, the system will reboot to its normal operational mode. You will then find Kernel Crash Dump file in the /var/crash directory :

ls /var/crash
linux-image-3.0.0-12-server.0.crash

Resources

Kernel Crash Dump is a vast topic that requires good knowledge of the linux kernel. You can find more information on the topic here :