Firmware assisted dump (fadump) HOWTO Introduction Firmware assisted dump is a new feature in the 3.4 mainline kernel supported only on powerpc architecture. The goal of firmware-assisted dump is to enable the dump of a crashed system, and to do so from a fully-reset system, and to minimize the total elapsed time until the system is back in production use. A complete documentation on implementation can be found at Documentation/powerpc/firmware-assisted-dump.txt in upstream linux kernel tree from 3.4 version and above. Please note that the firmware-assisted dump feature is only available on Power6 and above systems with recent firmware versions. Overview Fadump Fadump is a robust kernel crash dumping mechanism to get reliable kernel crash dump with assistance from firmware. This approach does not use kexec, instead firmware assists in booting the kdump kernel while preserving memory contents. Unlike kdump, the system is fully reset, and loaded with a fresh copy of the kernel. In particular, PCI and I/O devices are reinitialized and are in a clean, consistent state. This second kernel, often called a capture kernel, boots with very little memory and captures the dump image. The first kernel registers the sections of memory with the Power firmware for dump preservation during OS initialization. These registered sections of memory are reserved by the first kernel during early boot. When a system crashes, the Power firmware fully resets the system, preserves all the system memory contents, save the low memory (boot memory of size larger of 5% of system RAM or 256MB) of RAM to the previous registered region. It will also save system registers, and hardware PTE's. Fadump is supported only on ppc64 platform. The standard kernel and capture kernel are one and the same on ppc64. If you're reading this document, you should already have kexec-tools installed. If not, you install it via the following command: # yum install kexec-tools Fadump Operational Flow: Like kdump, fadump also exports the ELF formatted kernel crash dump through /proc/vmcore. Hence existing kdump infrastructure can be used to capture fadump vmcore. The idea is to keep the functionality transparent to end user. From user perspective there is no change in the way kdump init script works. However, unlike kdump, fadump does not pre-load kdump kernel and initrd into reserved memory, instead it always uses default OS initrd during second boot after crash. Hence, for fadump, we rebuild the new kdump initrd and replace it with default initrd. Before replacing existing default initrd we take a backup of original default initrd for user's reference. The dracut package has been enhanced to rebuild the default initrd with vmcore capture steps. The initrd image is rebuilt as per the configuration in /etc/kdump.conf file. The control flow of fadump works as follows: 01. System panics. 02. At the crash, kernel informs power firmware that kernel has crashed. 03. Firmware takes the control and reboots the entire system preserving only the memory (resets all other devices). 04. The reboot follows the normal booting process (non-kexec). 05. The boot loader loads the default kernel and initrd from /boot 06. The default initrd loads and runs /init 07. dracut-kdump.sh script present in fadump aware default initrd checks if '/proc/device-tree/rtas/ibm,kernel-dump' file exists before executing steps to capture vmcore. (This check will help to bypass the vmcore capture steps during normal boot process.) 09. Captures dump according to /etc/kdump.conf 10. Is dump capture successful (yes goto 12, no goto 11) 11. Perform the failure action specified in /etc/kdump.conf (The default failure action is reboot, if unspecified) 12. Perform the final action specified in /etc/kdump.conf (The default final action is reboot, if unspecified) How to configure fadump: Again, we assume if you're reading this document, you should already have kexec-tools installed. If not, you install it via the following command: # yum install kexec-tools Make the kernel to be configured with FADump as the default boot entry, if it isn't already: # grubby --set-default=/boot/vmlinuz- Boot into the kernel to be configured for FADump. To be able to do much of anything interesting in the way of debug analysis, you'll also need to install the kernel-debuginfo package, of the same arch as your running kernel, and the crash utility: # yum --enablerepo=\*debuginfo install kernel-debuginfo.$(uname -m) crash Next up, we need to modify some boot parameters to enable firmware assisted dump. With the help of grubby, it's very easy to append "fadump=on" to the end of your kernel boot parameters. To reserve the appropriate amount of memory for boot memory preservation, pass 'crashkernel=X' kernel cmdline parameter. For the recommended value of X, see 'FADump Memory Requirements' section. # grubby --args="fadump=on crashkernel=6G" --update-kernel=/boot/vmlinuz-`uname -r` By default, FADump reserved memory will be initialized as CMA area to make the memory available through CMA allocator on the production kernel. We can opt out of this, making reserved memory unavailable to production kernel, by booting the linux kernel with 'fadump=nocma' instead of 'fadump=on'. The term 'boot memory' means size of the low memory chunk that is required for a kernel to boot successfully when booted with restricted memory. By default, the boot memory size will be the larger of 5% of system RAM or 256MB. Alternatively, user can also specify boot memory size through boot parameter 'fadump_reserve_mem=' which will override the default calculated size. Use this option if default boot memory size is not sufficient for second kernel to boot successfully. After making said changes, reboot your system, so that the specified memory is reserved and left untouched by the normal system. Take note that the output of 'free -m' will show X MB less memory than without this parameter, which is expected. If you see OOM (Out Of Memory) error messages while loading capture kernel, then you should bump up the memory reservation size. Now that you've got that reserved memory region set up, you want to turn on the kdump init script: # systemctl enable kdump.service Then, start up kdump as well: # systemctl start kdump.service This should turn on the firmware assisted functionality in kernel by echo'ing 1 to /sys/kernel/fadump_registered, leaving the system ready to capture a vmcore upon crashing. For journaling filesystems like XFS an additional step is required to ensure bootloader does not pick the older initrd (without vmcore capture scripts): * If /boot is a separate partition, run the below commands as the root user, or as a user with CAP_SYS_ADMIN rights: # fsfreeze -f # fsfreeze -u * If /boot is not a separate partition, reboot the system. After reboot check if the kdump service is up and running with: # systemctl status kdump.service To test out whether FADump is configured properly, you can force-crash your system by echo'ing a 'c' into /proc/sysrq-trigger: # echo c > /proc/sysrq-trigger You should see some panic output, followed by the system reset and booting into fresh copy of kernel. When default initrd loads and runs /init, vmcore should be copied out to disk (by default, in /var/crash//vmcore), then the system rebooted back into your normal kernel. Once back to your normal kernel, you can use the previously installed crash kernel in conjunction with the previously installed kernel-debuginfo to perform postmortem analysis: # crash /usr/lib/debug/lib/modules/2.6.17-1.2621.el5/vmlinux /var/crash/2006-08-23-15:34/vmcore crash> bt and so on... Saving vmcore-dmesg.txt ----------------------- Kernel log bufferes are one of the most important information available in vmcore. Now before saving vmcore, kernel log bufferes are extracted from /proc/vmcore and saved into a file vmcore-dmesg.txt. After vmcore-dmesg.txt, vmcore is saved. Destination disk and directory for vmcore-dmesg.txt is same as vmcore. Note that kernel log buffers will not be available if dump target is raw device. FADump Memory Requirements: System Memory Recommended memory --------------------- ---------------------- 4 GB - 16 GB : 768 MB 16 GB - 64 GB : 1024 MB 64 GB - 128 GB : 2 GB 128 GB - 1 TB : 4 GB 1 TB - 2 TB : 6 GB 2 TB - 4 TB : 12 GB 4 TB - 8 TB : 20 GB 8 TB - 16 TB : 36 GB 16 TB - 32 TB : 64 GB 32 TB - 64 TB : 128 GB 64 TB & above : 180 GB Things to remember: 1) The memory required to boot capture Kernel is a moving target that depends on many factors like hardware attached to the system, kernel and modules in use, packages installed and services enabled, there is no one-size-fits-all. But the above recommendations are based on system memory. So, the above recommendations for FADump come with a few assumptions, based on available system memory, about the resources the system could have. So, please take the recommendations with a pinch of salt and remember to try capturing dump a few times to confirm that the system is configured successfully with dump capturing support. 2) Though the memory requirements for FADump seem high, this memory is not completely set aside but made available for userspace applications to use, through the CMA allocator. 3) As the same initrd is used for booting production kernel as well as capture kernel and with dump being captured in a restricted memory environment, few optimizations (like not inclding network dracut module, disabling multipath and such) are applied while building the initrd. In case, the production environment needs these optimizations to be avoided, dracut_args option in /etc/kdump.conf file could be leveraged. For example, if a user wishes for network module to be included in the initrd, adding the below entry in /etc/kdump.conf file and restarting kdump service would take care of it. dracut_args --add "network" 4) If FADump is configured to capture vmcore to a remote dump target using SSH or NFS protocol, the network interface is renamed to kdump- if is generic, for example, *eth#, or net#. This problem occurs because the vmcore capture scripts in the initial RAM disk (initrd) add the kdump- prefix to the network interface name to secure persistent naming. As the same initrd is used for production kernel boot, the interface name is changed for the production kernel too. Dump Triggering methods: This section talks about the various ways, other than a Kernel Panic, in which fadump can be triggered. The following methods assume that fadump is configured on your system, with the scripts enabled as described in the section above. 1) AltSysRq C FAdump can be triggered with the combination of the 'Alt','SysRq' and 'C' keyboard keys. Please refer to the following link for more details: https://fedoraproject.org/wiki/QA/Sysrq In addition, on PowerPC boxes, fadump can also be triggered via Hardware Management Console(HMC) using 'Ctrl', 'O' and 'C' keyboard keys. 2) Kernel OOPs If we want to generate a dump everytime the Kernel OOPses, we can achieve this by setting the 'Panic On OOPs' option as follows: # echo 1 > /proc/sys/kernel/panic_on_oops 3) PowerPC specific methods: On IBM PowerPC machines, issuing a soft reset invokes the XMON debugger(if XMON is configured). To configure XMON one needs to compile the kernel with the CONFIG_XMON and CONFIG_XMON_DEFAULT options, or by compiling with CONFIG_XMON and booting the kernel with xmon=on option. Following are the ways to remotely issue a soft reset on PowerPC boxes, which would drop you to XMON. Pressing a 'X' (capital alphabet X) followed by an 'Enter' here will trigger the dump. 3.1) HMC Hardware Management Console(HMC) available on Power4 and Power5 machines allow partitions to be reset remotely. This is specially useful in hang situations where the system is not accepting any keyboard inputs. Once you have HMC configured, the following steps will enable you to trigger fadump via a soft reset: On Power4 Using GUI * In the right pane, right click on the partition you wish to dump. * Select "Operating System->Reset". * Select "Soft Reset". * Select "Yes". Using HMC Commandline # reset_partition -m -p -t soft On Power5 Using GUI * In the right pane, right click on the partition you wish to dump. * Select "Restart Partition". * Select "Dump". * Select "OK". Using HMC Commandline # chsysstate -m -n -o dumprestart -r lpar 3.2) Blade Management Console for Blade Center To initiate a dump operation, go to Power/Restart option under "Blade Tasks" in the Blade Management Console. Select the corresponding blade for which you want to initate the dump and then click "Restart blade with NMI". This issues a system reset and invokes xmon debugger. Advanced Setups & Failure action: Kdump and fadump exhibit similar behavior in terms of setup & failure action. For fadump advanced setup related information see section "Advanced Setups" in "kexec-kdump-howto.txt" document. Refer to "Failure action" section in "kexec- kdump-howto.txt" document for fadump failure action related information. Compression and filtering Refer "Compression and filtering" section in "kexec-kdump-howto.txt" document. Compression and filtering are same for kdump & fadump. Notes on rootfs mount: Dracut is designed to mount rootfs by default. If rootfs mounting fails it will refuse to go on. So fadump leaves rootfs mounting to dracut currently. We make the assumtion that proper root= cmdline is being passed to dracut initramfs for the time being. If you need modify "KDUMP_COMMANDLINE=" in /etc/sysconfig/kdump, you will need to make sure that appropriate root= options are copied from /proc/cmdline. In general it is best to append command line options using "KDUMP_COMMANDLINE_APPEND=" instead of replacing the original command line completely. How to disable FADump: Remove "fadump=on"/"fadump=nocma" from kernel cmdline parameters OR replace it with "fadump=off" kernel cmdline parameter: # grubby --update-kernel=/boot/vmlinuz-`uname -r` --remove-args="fadump=on" or # grubby --update-kernel=/boot/vmlinuz-`uname -r` --remove-args="fadump=nocma" OR # grubby --update-kernel=/boot/vmlinuz-`uname -r` --args="fadump=off" If KDump is to be used as the dump capturing mechanism, update the crashkernel parameter (Else, remove "crashkernel=" parameter too, using grubby): # grubby --update-kernel=/boot/vmlinuz-$kver --args="crashkernl=auto" Reboot the system for the settings to take effect.