The kdump-tools package provides init scripts and configuration files to
use kdump.

It is as automated as can be at this point, but some manual configuration
may still be required.  Specifically:

1. Kernel Configuration
    You must boot a kernel that was configured with:
        CONFIG_KEXEC=y
        CONFIG_CRASH_DUMP=y
    For ia64, you also need:
        CONFIG_SPARSEMEM=y
    For the resulting dump to be useful, you probably also want:
        CONFIG_DEBUG_INFO=y

    If you use CONFIG_DISCONTIG, then you can only use makedumpfile level
    1 (omit zero pages).

2. Kdump Kernel
    You must have a kdump kernel, which is either relocatable or built to
    start at a non-default address (which you must also set in the kernel
    command-line parameter, see below).  Linux on ia64 is always
    relocatable; recent x86, x86_64, and powerpc kernels have an option
    for this:
        CONFIG_RELOCATABLE=y
    If your architecture does not support this option, you must build it
    with a non-default CONFIG_PHYSICAL_START option.

    If your boot kernel is relocatable, kdump-config will use it as the
    kdump kernel.  Otherwise, you will have to provide one.  Once you
    have a relocatable crash kernel, set KDUMP_KERNEL and if necessary
    KDUMP_INITRD in the /etc/default/kdump-tools file.

    The kdump kernel needs to be configured with:
        CONFIG_CRASH_DUMP=y

3. Kernel Command line parameter
    You must boot your kernel with a 'crashkernel=' command line parameter,
    for example:

      crashkernel=128M

    That will reserve 128 MB of memory for the kdump kernel to use in case
    of a crash.  See Documentation/kdump/kdump.txt in the Linux source for
    more advanced crashkernel parameter syntax.

    You may also want to add 'nmi_watchdog=1' on certain systems.

4. Debug Kernel
    You *should* have a debug kernel in order for makedumpfile to process
    the vmcore file.  Without a debug kernel, the transfer process is
    reduced to using "makedumpfile -d 1".   Options:
        A)  If /usr/lib/debug/boot/vmlinux-$(uname -r) exists,
	    kdump-tools will use that kernel.
        B)  Explicitly set DEBUG_KERNEL in /etc/default/kdump-tools to
            point to your debug kernel.
        C)  None of the above.  makedumpfile will still work, but your
            dumpfile will be larger and take longer to save to disk.

5. Local Configuration
   The /etc/default/kdump-tools file can be modified to reflect your 
   setup, if automatic detection fails, or if you need specific
   architectural settings.

6. Architectural considerations
   A) x86 && PAE && memory > 4 Gigabytes 
      will need to use KDUMP_KEXEC_ARGS="--elf64-core-headers"

   B) x86 and x86_64
      Some systems can take advantage of the nmi watchdog.  Add 
      nmi_watchdog=1 to the boot commandline to turn on the watchdog.
      The nmi interrupt will call panic if activated.

   C) ia64
      some systems may need KDUMP_KEXEC_ARGS="--noio".  Use this
      if the system hangs after a panic, but before the kdump kernel 
      begins to boot.