From Fedora Project Wiki

No edit summary
No edit summary
 
Line 8: Line 8:
#Apply changes and start kdump via service kdump restart
#Apply changes and start kdump via service kdump restart
|actions=
|actions=
{{admon/note|Download crash-simple.cmd and crash.cmd at|http://hongqing.fedorapeople.org/script/crash-simple.cmd http://hongqing.fedorapeople.org/script/crash.cmd}}
#Trigger crash via:<pre>echo c > /proc/sysrq-trigger</pre>
#Trigger crash via:<pre>echo c > /proc/sysrq-trigger</pre>
#Analyse vmcore:<pre>readelf -a /var/crash/xxx/vmcore</pre><pre>objdump -x /var/crash/xxx/vmcore</pre><pre>eu-readelf -a /var/crash/xxx/vmcore</pre>
#Analyse vmcore:<pre>readelf -a /var/crash/xxx/vmcore</pre><pre>objdump -x /var/crash/xxx/vmcore</pre><pre>eu-readelf -a /var/crash/xxx/vmcore</pre>

Latest revision as of 08:40, 26 March 2012

Description

Setup

  1. Install Fedora 17 on your machine
  2. After system installed, check if kexec-tools kernel-debuginfo and crash is installed via
    rpm -q kexec-tools kernel-debuginfo crash
    Or via yum
    yum install kexec-tools kernel-debuginfo crash
  3. Reserve crashkernel for kdump by grubby 
    grubby --args="crashkernl=128M" --update-kernel=$(grubby --default-kernel)
    Reboot system and check /proc/cmdline
    cat /proc/cmdline | grep "crashkernel"
  4. Edit /etc/kdump.conf, add this line:
     ext4 /dev/mapper/vg_dhcp6558-lv_root
    core_collector makedumpfile -E -d 31
  5. Apply changes and start kdump via service kdump restart

How to test

  1. Trigger crash via:
    echo c > /proc/sysrq-trigger
  2. Analyse vmcore:
    readelf -a /var/crash/xxx/vmcore
    objdump -x /var/crash/xxx/vmcore
    eu-readelf -a /var/crash/xxx/vmcore
  3. Use crash to analyse vmcore:
    crash -i crash-simple.cmd /usr/lib/debug/lib/modules/xxx/vmlinux /var/crash/xxx/vmcore
    crash -i crash.cmd /usr/lib/debug/lib/modules/xxx/vmlinux /var/crash/xxx/vmcore
  4. Use gdb to analyse vmcore:
    gdb -x gdb.cmd /usr/lib/debug/lib/modules/xxx/vmlinux /var/crash/xxx/vmcore

Expected Results

  1. Dump process could be monitered if console is connected
  2. System reboot successfully after dump vmcore finshed
  3. vmcore was found under path /var/crash/xxx/vmcore
  4. readelf should return 0, and no "warning" "warnings" found on output
    1. objdump should return 0
    2. eu-readelf should return 0
  5. No warning or error should be should found on crash output
  6. No warning or error should be should found on gdb output