sledp2-kdumpRecommended update for kdumpRecommended update for kdumpThis update fixes following problems in kdump configuration:
1. When taking a dump on machines that were booted with
the "mem" parameter, the dump is now only as large as the
specified memory, not as large as the full memory. That
affects only the i386 and the x86-64 architectures. (Update
of kexec-tools and kdump packages needed.)
2. Fix load of kdump kernel when KDUMP_KERNELVER="kdump"
on ppc64. (Only updating "kdump" is required for that fix.)
3. Support "kernel-kdumppae" to take dumps on machines
that have more than 4 GiB of physical memory but are 32 bit
(or use a 32 bit kernel). That update contains only the
support in "kdump" package (init scripts), not the actual
"kernel-kdumppae" kernel.
4. Replace KDUMP_KERNELVER="kdump" with KDUMP_KERNELVER=""
in /etc/sysconfig/kdump when installing the updated
package. That is required to use a "kernel-kdumppae"
automatically without editing the configuration file once
it gots released. (Only update of "kdump" is required for
that fix.)
5. Set KDUMP_KERNELVER="" on i386 instead of
KDUMP_KERNELVER="kdump". That affects only new
installations of the "kdump" package. For updates, the
change (4) fixes the problem.
6. Try also $(uname -i) to find the makedumpfile.config in
/usr/src/linux-${version}-obj/${arch}/${flavour}. The
difference between $(uname -i) and $(uname -m) is that it's
"i386" always on i386 and $(uname -m) varies between "i586"
and "i686".
7. Implement "try-restart" command and use that
implementation also for "condrestart". The difference
between "restart" and "try-restart" is that "try-restart"
checks if the service was already active. If not, does
nothing. That update affects only the "kdump" package and
was required for fix (7).
8. Add udev rules to reload kdump configuration on memory
or CPU hotplug. This is required to get also a dump of the
"new" memory. The so-called ELF core headers (i.e. the
description of the memory that is dumped) is generated on
kdump load, not when the actual crash occurs. (That fix
also affects only "kdump" package.)
9. Panic on unknown NMI automatically. Some machines have
an "NMI" button that can be used to get a dump even if
SysRq doesn't work any more and the machine has locked-up.
(That fix only affects "kdump" package.)
10. Print stderr of kexec invocation. Previously, warnings
printed by the "kexec" command have been silently ignored
when the return value of the command was 0 (success).
For all that reasons, it's highly recommended to install
that update.
This update fixes following problems in kdump configuration:
1. When taking a dump on machines that were booted with
the "mem" parameter, the dump is now only as large as the
specified memory, not as large as the full memory. That
affects only the i386 and the x86-64 architectures. (Update
of kexec-tools and kdump packages needed.)
2. Fix load of kdump kernel when KDUMP_KERNELVER="kdump"
on ppc64. (Only updating "kdump" is required for that fix.)
3. Support "kernel-kdumppae" to take dumps on machines
that have more than 4 GiB of physical memory but are 32 bit
(or use a 32 bit kernel). That update contains only the
support in "kdump" package (init scripts), not the actual
"kernel-kdumppae" kernel.
4. Replace KDUMP_KERNELVER="kdump" with KDUMP_KERNELVER=""
in /etc/sysconfig/kdump when installing the updated
package. That is required to use a "kernel-kdumppae"
automatically without editing the configuration file once
it gots released. (Only update of "kdump" is required for
that fix.)
5. Set KDUMP_KERNELVER="" on i386 instead of
KDUMP_KERNELVER="kdump". That affects only new
installations of the "kdump" package. For updates, the
change (4) fixes the problem.
6. Try also $(uname -i) to find the makedumpfile.config in
/usr/src/linux-${version}-obj/${arch}/${flavour}. The
difference between $(uname -i) and $(uname -m) is that it's
"i386" always on i386 and $(uname -m) varies between "i586"
and "i686".
7. Implement "try-restart" command and use that
implementation also for "condrestart". The difference
between "restart" and "try-restart" is that "try-restart"
checks if the service was already active. If not, does
nothing. That update affects only the "kdump" package and
was required for fix (7).
8. Add udev rules to reload kdump configuration on memory
or CPU hotplug. This is required to get also a dump of the
"new" memory. The so-called ELF core headers (i.e. the
description of the memory that is dumped) is generated on
kdump load, not when the actual crash occurs. (That fix
also affects only "kdump" package.)
9. Panic on unknown NMI automatically. Some machines have
an "NMI" button that can be used to get a dump even if
SysRq doesn't work any more and the machine has locked-up.
(That fix only affects "kdump" package.)
10. Print stderr of kexec invocation. Previously, warnings
printed by the "kexec" command have been silently ignored
when the return value of the command was 0 (success).
For all that reasons, it's highly recommended to install
that update.
recommendedkdumpx86_64b2c7fa6a7fdf0736dab89a81a6d5a56358315fa9