================================================================ Documentation for Kdump - The kexec-based Crash Dumping Solution ================================================================ This document includes overview, setup and installation, and analysis information. Overview ======== Kdump uses kexec to quickly boot to a dump-capture kernel whenever a dump of the system kernel's memory needs to be taken (for example, when the system panics). The system kernel's memory image is preserved across the reboot and is accessible to the dump-capture kernel. You can use common Linux commands, such as cp and scp, to copy the memory image to a dump file on the local disk, or across the network to a remote system. Kdump and kexec are currently supported on the x86, x86_64, and ppc64 architectures. When the system kernel boots, it reserves a small section of memory for the dump-capture kernel. This ensures that ongoing Direct Memory Access (DMA) from the system kernel does not corrupt the dump-capture kernel. The kexec -p command loads the dump-capture kernel into this reserved memory. On x86 machines, the first 640 KB of physical memory is needed to boot, regardless of where the kernel loads. Therefore, kexec backs up this region just before rebooting into the dump-capture kernel. All of the necessary information about the system kernel's core image is encoded in the ELF format, and stored in a reserved area of memory before a crash. The physical address of the start of the ELF header is passed to the dump-capture kernel through the elfcorehdr= boot parameter. With the dump-capture kernel, you can access the memory image, or "old memory," in two ways: - Through a /dev/oldmem device interface. A capture utility can read the device file and write out the memory in raw format. This is a raw dump of memory. Analysis and capture tools must be intelligent enough to determine where to look for the right information. - Through /proc/vmcore. This exports the dump as an ELF-format file that you can write out using file copy commands such as cp or scp. Further, you can use analysis tools such as the GNU Debugger (GDB) and the Crash tool to debug the dump file. This method ensures that the dump pages are correctly ordered. Setup and Installation ====================== Install kexec-tools and the Kdump patch --------------------------------------- 1) Login as the root user. 2) Download the kexec-tools user-space package from the following URL: http://www.xmission.com/~ebiederm/files/kexec/kexec-tools-1.101.tar.gz 3) Unpack the tarball with the tar command, as follows: tar xvpzf kexec-tools-1.101.tar.gz 4) Download the latest consolidated Kdump patch from the following URL: http://lse.sourceforge.net/kdump/ (This location is being used until all the user-space Kdump patches are integrated with the kexec-tools package.) 5) Change to the kexec-tools-1.101 directory, as follows: cd kexec-tools-1.101 6) Apply the consolidated patch to the kexec-tools-1.101 source tree with the patch command, as follows. (Modify the path to the downloaded patch as necessary.) patch -p1 < /path-to-kdump-patch/kexec-tools-1.101-kdump.patch 7) Configure the package, as follows: ./configure 8) Compile the package, as follows: make 9) Install the package, as follows: make install Download and build the system and dump-capture kernels ------------------------------------------------------ Download the mainline (vanilla) kernel source code (2.6.13-rc1 or newer) from http://www.kernel.org. Two kernels must be built: a system kernel and a dump-capture kernel. Use the following steps to configure these kernels with the necessary kexec and Kdump features: System kernel ------------- 1) Enable "kexec system call" in "Processor type and features." CONFIG_KEXEC=y 2) Enable "sysfs file system support" in "Filesystem" -> "Pseudo filesystems." This is usually enabled by default. CONFIG_SYSFS=y Note that "sysfs file system support" might not appear in the "Pseudo filesystems" menu if "Configure standard kernel features (for small systems)" is not enabled in "General Setup." In this case, check the .config file itself to ensure that sysfs is turned on, as follows: grep 'CONFIG_SYSFS' .config 3) Enable "Compile the kernel with debug info" in "Kernel hacking." CONFIG_DEBUG_INFO=Y This causes the kernel to be built with debug symbols. The dump analysis tools require a vmlinux with debug symbols in order to read and analyze a dump file. 4) Make and install the kernel and its modules. Update the boot loader (such as grub, yaboot, or lilo) configuration files as necessary. 5) Boot the system kernel with the boot parameter "crashkernel=Y@X", where Y specifies how much memory to reserve for the dump-capture kernel and X specifies the beginning of this reserved memory. For example, "crashkernel=64M@16M" tells the system kernel to reserve 64 MB of memory starting at physical address 0x01000000 for the dump-capture kernel. On x86 and x86_64, use "crashkernel=64M@16M". On ppc64, use "crashkernel=128M@32M". The dump-capture kernel ----------------------- 1) Under "General setup," append "-kdump" to the current string in "Local version." 2) On x86, enable high memory support under "Processor type and features": CONFIG_HIGHMEM64G=y or CONFIG_HIGHMEM4G 3) On x86 and x86_64, disable symmetric multi-processing support under "Processor type and features": CONFIG_SMP=n (If CONFIG_SMP=y, then specify maxcpus=1 on the kernel command line when loading the dump-capture kernel, see section "Load the Dump-capture Kernel".) 4) On ppc64, disable NUMA support and enable EMBEDDED support: CONFIG_NUMA=n CONFIG_EMBEDDED=y CONFIG_EEH=N for the dump-capture kernel 5) Enable "kernel crash dumps" support under "Processor type and features": CONFIG_CRASH_DUMP=y 6) Use a suitable value for "Physical address where the kernel is loaded" (under "Processor type and features"). This only appears when "kernel crash dumps" is enabled. By default this value is 0x1000000 (16MB). It should be the same as X in the "crashkernel=Y@X" boot parameter discussed above. On x86 and x86_64, use "CONFIG_PHYSICAL_START=0x1000000". On ppc64 the value is automatically set at 32MB when CONFIG_CRASH_DUMP is set. 6) Optionally enable "/proc/vmcore support" under "Filesystems" -> "Pseudo filesystems". CONFIG_PROC_VMCORE=y (CONFIG_PROC_VMCORE is set by default when CONFIG_CRASH_DUMP is selected.) 7) Make and install the kernel and its modules. DO NOT add this kernel to the boot loader configuration files. Load the Dump-capture Kernel ============================ After booting to the system kernel, load the dump-capture kernel using the following command: kexec -p \ --initrd= --args-linux \ --append="root= init 1 irqpoll" Notes on loading the dump-capture kernel: * must be a vmlinux image (that is, an uncompressed ELF image). bzImage does not work at this time. * By default, the ELF headers are stored in ELF64 format to support systems with more than 4GB memory. The --elf32-core-headers option can be used to force the generation of ELF32 headers. This is necessary because GDB currently cannot open vmcore files with ELF64 headers on 32-bit systems. ELF32 headers can be used on non-PAE systems (that is, less than 4GB of memory). * The "irqpoll" boot parameter reduces driver initialization failures due to shared interrupts in the dump-capture kernel. * You must specify in the format corresponding to the root device name in the output of mount command. * "init 1" boots the dump-capture kernel into single-user mode without networking. If you want networking, use "init 3." Kernel Panic ============ After successfully loading the dump-capture kernel as previously described, the system will reboot into the dump-capture kernel if a system crash is triggered. Trigger points are located in panic(), die(), die_nmi() and in the sysrq handler (ALT-SysRq-c). The following conditions will execute a crash trigger point: If a hard lockup is detected and "NMI watchdog" is configured, the system will boot into the dump-capture kernel ( die_nmi() ). If die() is called, and it happens to be a thread with pid 0 or 1, or die() is called inside interrupt context or die() is called and panic_on_oops is set, the system will boot into the dump-capture kernel. On powererpc systems when a soft-reset is generated, die() is called by all cpus and the system will boot into the dump-capture kernel. For testing purposes, you can trigger a crash by using "ALT-SysRq-c", "echo c > /proc/sysrq-trigger or write a module to force the panic. Write Out the Dump File ======================= After the dump-capture kernel is booted, write out the dump file with the following command: cp /proc/vmcore You can also access dumped memory as a /dev/oldmem device for a linear and raw view. To create the device, use the following command: mknod /dev/oldmem c 1 12 Use the dd command with suitable options for count, bs, and skip to access specific portions of the dump. To see the entire memory, use the following command: dd if=/dev/oldmem of=oldmem.001 Analysis ======== Before analyzing the dump image, you should reboot into a stable kernel. You can do limited analysis using GDB on the dump file copied out of /proc/vmcore. Use the debug vmlinux built with -g and run the following command: gdb vmlinux Stack trace for the task on processor 0, register display, and memory display work fine. Note: GDB cannot analyze core files generated in ELF64 format for x86. On systems with a maximum of 4GB of memory, you can generate ELF32-format headers using the --elf32-core-headers kernel option on the dump kernel. You can also use the Crash utility to analyze dump files in Kdump format. Crash is available on Dave Anderson's site at the following URL: http://people.redhat.com/~anderson/ To Do ===== 1) Provide a kernel pages filtering mechanism, so core file size is not extreme on systems with huge memory banks. 2) Relocatable kernel can help in maintaining multiple kernels for crash_dump, and the same kernel as the system kernel can be used to capture the dump. Contact ======= Vivek Goyal (vgoyal@in.ibm.com) Maneesh Soni (maneesh@in.ibm.com) Trademark ========= Linux is a trademark of Linus Torvalds in the United States, other countries, or both.