1 ================================================================
2 Documentation for Kdump - The kexec-based Crash Dumping Solution
3 ================================================================
5 This document includes overview, setup, installation, and analysis
11 Kdump uses kexec to quickly boot to a dump-capture kernel whenever a
12 dump of the system kernel's memory needs to be taken (for example, when
13 the system panics). The system kernel's memory image is preserved across
14 the reboot and is accessible to the dump-capture kernel.
16 You can use common commands, such as cp, scp or makedumpfile to copy
17 the memory image to a dump file on the local disk, or across the network
20 Kdump and kexec are currently supported on the x86, x86_64, ppc64, ia64,
21 s390x, arm and arm64 architectures.
23 When the system kernel boots, it reserves a small section of memory for
24 the dump-capture kernel. This ensures that ongoing Direct Memory Access
25 (DMA) from the system kernel does not corrupt the dump-capture kernel.
26 The kexec -p command loads the dump-capture kernel into this reserved
29 On x86 machines, the first 640 KB of physical memory is needed for boot,
30 regardless of where the kernel loads. For simpler handling, the whole
31 low 1M is reserved to avoid any later kernel or device driver writing
32 data into this area. Like this, the low 1M can be reused as system RAM
33 by kdump kernel without extra handling.
35 On PPC64 machines first 32KB of physical memory is needed for booting
36 regardless of where the kernel is loaded and to support 64K page size
37 kexec backs up the first 64KB memory.
39 For s390x, when kdump is triggered, the crashkernel region is exchanged
40 with the region [0, crashkernel region size] and then the kdump kernel
41 runs in [0, crashkernel region size]. Therefore no relocatable kernel is
44 All of the necessary information about the system kernel's core image is
45 encoded in the ELF format, and stored in a reserved area of memory
46 before a crash. The physical address of the start of the ELF header is
47 passed to the dump-capture kernel through the elfcorehdr= boot
48 parameter. Optionally the size of the ELF header can also be passed
49 when using the elfcorehdr=[size[KMG]@]offset[KMG] syntax.
51 With the dump-capture kernel, you can access the memory image through
52 /proc/vmcore. This exports the dump as an ELF-format file that you can
53 write out using file copy commands such as cp or scp. You can also use
54 makedumpfile utility to analyze and write out filtered contents with
55 options, e.g with '-d 31' it will only write out kernel data. Further,
56 you can use analysis tools such as the GNU Debugger (GDB) and the Crash
57 tool to debug the dump file. This method ensures that the dump pages are
60 Setup and Installation
61 ======================
66 1) Login as the root user.
68 2) Download the kexec-tools user-space package from the following URL:
70 http://kernel.org/pub/linux/utils/kernel/kexec/kexec-tools.tar.gz
72 This is a symlink to the latest version.
74 The latest kexec-tools git tree is available at:
76 - git://git.kernel.org/pub/scm/utils/kernel/kexec/kexec-tools.git
77 - http://www.kernel.org/pub/scm/utils/kernel/kexec/kexec-tools.git
79 There is also a gitweb interface available at
80 http://www.kernel.org/git/?p=utils/kernel/kexec/kexec-tools.git
82 More information about kexec-tools can be found at
83 http://horms.net/projects/kexec/
85 3) Unpack the tarball with the tar command, as follows::
87 tar xvpzf kexec-tools.tar.gz
89 4) Change to the kexec-tools directory, as follows::
91 cd kexec-tools-VERSION
93 5) Configure the package, as follows::
97 6) Compile the package, as follows::
101 7) Install the package, as follows::
106 Build the system and dump-capture kernels
107 -----------------------------------------
108 There are two possible methods of using Kdump.
110 1) Build a separate custom dump-capture kernel for capturing the
113 2) Or use the system kernel binary itself as dump-capture kernel and there is
114 no need to build a separate dump-capture kernel. This is possible
115 only with the architectures which support a relocatable kernel. As
116 of today, i386, x86_64, ppc64, ia64, arm and arm64 architectures support
119 Building a relocatable kernel is advantageous from the point of view that
120 one does not have to build a second kernel for capturing the dump. But
121 at the same time one might want to build a custom dump capture kernel
122 suitable to his needs.
124 Following are the configuration setting required for system and
125 dump-capture kernels for enabling kdump support.
127 System kernel config options
128 ----------------------------
130 1) Enable "kexec system call" or "kexec file based system call" in
131 "Processor type and features."::
133 CONFIG_KEXEC=y or CONFIG_KEXEC_FILE=y
135 And both of them will select KEXEC_CORE::
139 Subsequently, CRASH_CORE is selected by KEXEC_CORE::
143 2) Enable "sysfs file system support" in "Filesystem" -> "Pseudo
144 filesystems." This is usually enabled by default::
148 Note that "sysfs file system support" might not appear in the "Pseudo
149 filesystems" menu if "Configure standard kernel features (for small
150 systems)" is not enabled in "General Setup." In this case, check the
151 .config file itself to ensure that sysfs is turned on, as follows::
153 grep 'CONFIG_SYSFS' .config
155 3) Enable "Compile the kernel with debug info" in "Kernel hacking."::
159 This causes the kernel to be built with debug symbols. The dump
160 analysis tools require a vmlinux with debug symbols in order to read
161 and analyze a dump file.
163 Dump-capture kernel config options (Arch Independent)
164 -----------------------------------------------------
166 1) Enable "kernel crash dumps" support under "Processor type and
171 2) Enable "/proc/vmcore support" under "Filesystems" -> "Pseudo filesystems"::
175 (CONFIG_PROC_VMCORE is set by default when CONFIG_CRASH_DUMP is selected.)
177 Dump-capture kernel config options (Arch Dependent, i386 and x86_64)
178 --------------------------------------------------------------------
180 1) On i386, enable high memory support under "Processor type and
189 2) With CONFIG_SMP=y, usually nr_cpus=1 need specified on the kernel
190 command line when loading the dump-capture kernel because one
191 CPU is enough for kdump kernel to dump vmcore on most of systems.
193 However, you can also specify nr_cpus=X to enable multiple processors
194 in kdump kernel. In this case, "disable_cpu_apicid=" is needed to
195 tell kdump kernel which cpu is 1st kernel's BSP. Please refer to
196 admin-guide/kernel-parameters.txt for more details.
198 With CONFIG_SMP=n, the above things are not related.
200 3) A relocatable kernel is suggested to be built by default. If not yet,
201 enable "Build a relocatable kernel" support under "Processor type and
206 4) Use a suitable value for "Physical address where the kernel is
207 loaded" (under "Processor type and features"). This only appears when
208 "kernel crash dumps" is enabled. A suitable value depends upon
209 whether kernel is relocatable or not.
211 If you are using a relocatable kernel use CONFIG_PHYSICAL_START=0x100000
212 This will compile the kernel for physical address 1MB, but given the fact
213 kernel is relocatable, it can be run from any physical address hence
214 kexec boot loader will load it in memory region reserved for dump-capture
217 Otherwise it should be the start of memory region reserved for
218 second kernel using boot parameter "crashkernel=Y@X". Here X is
219 start of memory region reserved for dump-capture kernel.
220 Generally X is 16MB (0x1000000). So you can set
221 CONFIG_PHYSICAL_START=0x1000000
223 5) Make and install the kernel and its modules. DO NOT add this kernel
224 to the boot loader configuration files.
226 Dump-capture kernel config options (Arch Dependent, ppc64)
227 ----------------------------------------------------------
229 1) Enable "Build a kdump crash kernel" support under "Kernel" options::
233 2) Enable "Build a relocatable kernel" support::
237 Make and install the kernel and its modules.
239 Dump-capture kernel config options (Arch Dependent, ia64)
240 ----------------------------------------------------------
242 - No specific options are required to create a dump-capture kernel
243 for ia64, other than those specified in the arch independent section
244 above. This means that it is possible to use the system kernel
245 as a dump-capture kernel if desired.
247 The crashkernel region can be automatically placed by the system
248 kernel at runtime. This is done by specifying the base address as 0,
249 or omitting it all together::
257 Dump-capture kernel config options (Arch Dependent, arm)
258 ----------------------------------------------------------
260 - To use a relocatable kernel,
261 Enable "AUTO_ZRELADDR" support under "Boot" options::
265 Dump-capture kernel config options (Arch Dependent, arm64)
266 ----------------------------------------------------------
268 - Please note that kvm of the dump-capture kernel will not be enabled
269 on non-VHE systems even if it is configured. This is because the CPU
270 will not be reset to EL2 on panic.
273 ===========================
274 1) crashkernel=size@offset
276 Here 'size' specifies how much memory to reserve for the dump-capture kernel
277 and 'offset' specifies the beginning of this reserved memory. For example,
278 "crashkernel=64M@16M" tells the system kernel to reserve 64 MB of memory
279 starting at physical address 0x01000000 (16MB) for the dump-capture kernel.
281 The crashkernel region can be automatically placed by the system
282 kernel at run time. This is done by specifying the base address as 0,
283 or omitting it all together::
291 If the start address is specified, note that the start address of the
292 kernel will be aligned to a value (which is Arch dependent), so if the
293 start address is not then any space below the alignment point will be
296 2) range1:size1[,range2:size2,...][@offset]
298 While the "crashkernel=size[@offset]" syntax is sufficient for most
299 configurations, sometimes it's handy to have the reserved memory dependent
300 on the value of System RAM -- that's mostly for distributors that pre-setup
301 the kernel command line to avoid a unbootable system after some memory has
302 been removed from the machine.
306 crashkernel=<range1>:<size1>[,<range2>:<size2>,...][@offset]
311 crashkernel=512M-2G:64M,2G-:128M
315 1) if the RAM is smaller than 512M, then don't reserve anything
316 (this is the "rescue" case)
317 2) if the RAM size is between 512M and 2G (exclusive), then reserve 64M
318 3) if the RAM size is larger than 2G, then reserve 128M
320 3) crashkernel=size,high and crashkernel=size,low
322 If memory above 4G is preferred, crashkernel=size,high can be used to
323 fulfill that. With it, physical memory is allowed to be allocated from top,
324 so could be above 4G if system has more than 4G RAM installed. Otherwise,
325 memory region will be allocated below 4G if available.
327 When crashkernel=X,high is passed, kernel could allocate physical memory
328 region above 4G, low memory under 4G is needed in this case. There are
329 three ways to get low memory:
331 1) Kernel will allocate at least 256M memory below 4G automatically
332 if crashkernel=Y,low is not specified.
333 2) Let user specify low memory size instead.
334 3) Specified value 0 will disable low memory allocation::
338 Boot into System Kernel
339 -----------------------
340 1) Update the boot loader (such as grub, yaboot, or lilo) configuration
343 2) Boot the system kernel with the boot parameter "crashkernel=Y@X".
345 On x86 and x86_64, use "crashkernel=Y[@X]". Most of the time, the
346 start address 'X' is not necessary, kernel will search a suitable
347 area. Unless an explicit start address is expected.
349 On ppc64, use "crashkernel=128M@32M".
351 On ia64, 256M@256M is a generous value that typically works.
352 The region may be automatically placed on ia64, see the
353 dump-capture kernel config option notes above.
354 If use sparse memory, the size should be rounded to GRANULE boundaries.
356 On s390x, typically use "crashkernel=xxM". The value of xx is dependent
357 on the memory consumption of the kdump system. In general this is not
358 dependent on the memory size of the production system.
360 On arm, the use of "crashkernel=Y@X" is no longer necessary; the
361 kernel will automatically locate the crash kernel image within the
362 first 512MB of RAM if X is not given.
364 On arm64, use "crashkernel=Y[@X]". Note that the start address of
365 the kernel, X if explicitly specified, must be aligned to 2MiB (0x200000).
367 Load the Dump-capture Kernel
368 ============================
370 After booting to the system kernel, dump-capture kernel needs to be
373 Based on the architecture and type of image (relocatable or not), one
374 can choose to load the uncompressed vmlinux or compressed bzImage/vmlinuz
375 of dump-capture kernel. Following is the summary.
379 - Use bzImage/vmlinuz if kernel is relocatable.
380 - Use vmlinux if kernel is not relocatable.
388 - Use vmlinux or vmlinuz.gz
392 - Use image or bzImage
400 - Use vmlinux or Image
402 If you are using an uncompressed vmlinux image then use following command
403 to load dump-capture kernel::
405 kexec -p <dump-capture-kernel-vmlinux-image> \
406 --initrd=<initrd-for-dump-capture-kernel> --args-linux \
407 --append="root=<root-dev> <arch-specific-options>"
409 If you are using a compressed bzImage/vmlinuz, then use following command
410 to load dump-capture kernel::
412 kexec -p <dump-capture-kernel-bzImage> \
413 --initrd=<initrd-for-dump-capture-kernel> \
414 --append="root=<root-dev> <arch-specific-options>"
416 If you are using a compressed zImage, then use following command
417 to load dump-capture kernel::
419 kexec --type zImage -p <dump-capture-kernel-bzImage> \
420 --initrd=<initrd-for-dump-capture-kernel> \
421 --dtb=<dtb-for-dump-capture-kernel> \
422 --append="root=<root-dev> <arch-specific-options>"
424 If you are using an uncompressed Image, then use following command
425 to load dump-capture kernel::
427 kexec -p <dump-capture-kernel-Image> \
428 --initrd=<initrd-for-dump-capture-kernel> \
429 --append="root=<root-dev> <arch-specific-options>"
431 Please note, that --args-linux does not need to be specified for ia64.
432 It is planned to make this a no-op on that architecture, but for now
435 Following are the arch specific command line options to be used while
436 loading dump-capture kernel.
438 For i386, x86_64 and ia64:
440 "1 irqpoll nr_cpus=1 reset_devices"
444 "1 maxcpus=1 noirqdistrib reset_devices"
448 "1 nr_cpus=1 cgroup_disable=memory"
452 "1 maxcpus=1 reset_devices"
456 "1 nr_cpus=1 reset_devices"
458 Notes on loading the dump-capture kernel:
460 * By default, the ELF headers are stored in ELF64 format to support
461 systems with more than 4GB memory. On i386, kexec automatically checks if
462 the physical RAM size exceeds the 4 GB limit and if not, uses ELF32.
463 So, on non-PAE systems, ELF32 is always used.
465 The --elf32-core-headers option can be used to force the generation of ELF32
466 headers. This is necessary because GDB currently cannot open vmcore files
467 with ELF64 headers on 32-bit systems.
469 * The "irqpoll" boot parameter reduces driver initialization failures
470 due to shared interrupts in the dump-capture kernel.
472 * You must specify <root-dev> in the format corresponding to the root
473 device name in the output of mount command.
475 * Boot parameter "1" boots the dump-capture kernel into single-user
476 mode without networking. If you want networking, use "3".
478 * We generally don't have to bring up a SMP kernel just to capture the
479 dump. Hence generally it is useful either to build a UP dump-capture
480 kernel or specify maxcpus=1 option while loading dump-capture kernel.
481 Note, though maxcpus always works, you had better replace it with
482 nr_cpus to save memory if supported by the current ARCH, such as x86.
484 * You should enable multi-cpu support in dump-capture kernel if you intend
485 to use multi-thread programs with it, such as parallel dump feature of
486 makedumpfile. Otherwise, the multi-thread program may have a great
487 performance degradation. To enable multi-cpu support, you should bring up an
488 SMP dump-capture kernel and specify maxcpus/nr_cpus, disable_cpu_apicid=[X]
489 options while loading it.
491 * For s390x there are two kdump modes: If a ELF header is specified with
492 the elfcorehdr= kernel parameter, it is used by the kdump kernel as it
493 is done on all other architectures. If no elfcorehdr= kernel parameter is
494 specified, the s390x kdump kernel dynamically creates the header. The
495 second mode has the advantage that for CPU and memory hotplug, kdump has
496 not to be reloaded with kexec_load().
498 * For s390x systems with many attached devices the "cio_ignore" kernel
499 parameter should be used for the kdump kernel in order to prevent allocation
500 of kernel memory for devices that are not relevant for kdump. The same
501 applies to systems that use SCSI/FCP devices. In that case the
502 "allow_lun_scan" zfcp module parameter should be set to zero before
503 setting FCP devices online.
508 After successfully loading the dump-capture kernel as previously
509 described, the system will reboot into the dump-capture kernel if a
510 system crash is triggered. Trigger points are located in panic(),
511 die(), die_nmi() and in the sysrq handler (ALT-SysRq-c).
513 The following conditions will execute a crash trigger point:
515 If a hard lockup is detected and "NMI watchdog" is configured, the system
516 will boot into the dump-capture kernel ( die_nmi() ).
518 If die() is called, and it happens to be a thread with pid 0 or 1, or die()
519 is called inside interrupt context or die() is called and panic_on_oops is set,
520 the system will boot into the dump-capture kernel.
522 On powerpc systems when a soft-reset is generated, die() is called by all cpus
523 and the system will boot into the dump-capture kernel.
525 For testing purposes, you can trigger a crash by using "ALT-SysRq-c",
526 "echo c > /proc/sysrq-trigger" or write a module to force the panic.
528 Write Out the Dump File
529 =======================
531 After the dump-capture kernel is booted, write out the dump file with
532 the following command::
534 cp /proc/vmcore <dump-file>
536 You can also use makedumpfile utility to write out the dump file
537 with specified options to filter out unwanted contents, e.g::
539 makedumpfile -l --message-level 1 -d 31 /proc/vmcore <dump-file>
544 Before analyzing the dump image, you should reboot into a stable kernel.
546 You can do limited analysis using GDB on the dump file copied out of
547 /proc/vmcore. Use the debug vmlinux built with -g and run the following
550 gdb vmlinux <dump-file>
552 Stack trace for the task on processor 0, register display, and memory
555 Note: GDB cannot analyze core files generated in ELF64 format for x86.
556 On systems with a maximum of 4GB of memory, you can generate
557 ELF32-format headers using the --elf32-core-headers kernel option on the
560 You can also use the Crash utility to analyze dump files in Kdump
561 format. Crash is available at the following URL:
563 https://github.com/crash-utility/crash
565 Crash document can be found at:
566 https://crash-utility.github.io/
568 Trigger Kdump on WARN()
569 =======================
571 The kernel parameter, panic_on_warn, calls panic() in all WARN() paths. This
572 will cause a kdump to occur at the panic() call. In cases where a user wants
573 to specify this during runtime, /proc/sys/kernel/panic_on_warn can be set to 1
574 to achieve the same behaviour.
576 Trigger Kdump on add_taint()
577 ============================
579 The kernel parameter panic_on_taint facilitates a conditional call to panic()
580 from within add_taint() whenever the value set in this bitmask matches with the
581 bit flag being set by add_taint().
582 This will cause a kdump to occur at the add_taint()->panic() call.
587 - kexec@lists.infradead.org
592 .. include:: gdbmacros.txt