Kdump Test Results

Following are the test reports for kexec/kdump testing, community has been doing so far. Please click at the respective fields for details like cpuinfo and lspci, test procedure, dump analysis etc. As of now the tests are conducted on i386 machines with different devices. The current issues are listed at the end. This is all work in progress and it will get appended and improved regularly.

Please help in fleshing up this page by sending kexec/kdump test results to Maneesh Soni or Vivek Goyal. The possible steps for testing kexec/kdump can be seen at this link Kexec/Kdump test plan

Also log any bugs seen in kexec/kdump/kexec-tools at OSDL Bugme . Please let us know the system details like cpuinfo,lspci, console/boot log etc. also.

Dump Analysis tips:

Please find here a link to useful set of gdb macros (user defined commands) which can be used to aid dump analysis from Alexander Nyberg.

Kexec/kdump Basic Test Reports:

Test Reports for i386 architecture:

Test Reports for x86_64 architecture:

Test Reports for ppc64 architecture:

Kexec/kdump Test reports using LKDTT test tool:

DTT (Dump Test Tool) is a test suite that evaluates the reliability of kernel crash dump capturing mechanisms for Linux by precisely recreating crash scenarios that take into account both the state of HW and the load conditions of the system. For more info http://www.lkdtt.sourceforge.net/

Test Reports for i386 architecture:

Test Reports for x86_64 architecture:

Test Reports for PPC64 architecture:

Issues:

1. Device initialisation in dump capture kernel. Intermittant failures are seen with devices like IBM servraid, AIC7xxx. USB initialisation issues have also been reported.
It is seen that building dump capture kernel with CONFIG_X86_UP_APIC=y and CONFIG_X86_UP_IOAPIC=y reduces the chances for hitting "IRQ nobody cared problem

2. On some machines kexec-on-panic is not able to boot SMP kernel as dump capture kernel.
As we neednot do anything else apart from copying the dump, the dump capture kernel can be built with CONFIG_SMP=N to work around this problem

3. Cannot boot any kernel with maxcpus=1 since 2.6.12-rc2-mm1
Same work around of building the dump capture kernel with CONFIG_SMP=N can be used here also

4. GDB related issues:
a) Cannot open dump files more than 2G
The Later versions of gdb like (GNU gdb Red Hat Linux (6.1post-1.20040607.62rh)) can open more than 2GB files

b) "bt" Stack traces doesn't terminate and continues to dump
Use "bt [number of frames to print]" to limit the output of bt

c) ELF64 formated dumps couldn't be analysied on IA32 systems

d) Same stack trace for all the CPUs in idle state.

5. Second kernel boots fine but nothing is seen at console, if vga=791 is specified in the kernel boot parameter. Though X also comes up fine.

6. Second kernel (vmlinux format) cannot be loaded with initrd while using -p (kexec-on-panic) option.
The latest kdump patch for kexec-tools has fixed this problem.

7. Second kernel fails to reboot in first kernel has KGDB enabled.