关于kdump的说明文档

Based on kernel version 3.1. Page generated on 2011-11-02 23:35 EST.


1 ================================================================
2 Documentation for Kdump - The kexec-based Crash Dumping Solution
3 ================================================================
4
5 This document includes overview, setup and installation, and analysis
6 information.
7
8 Overview
9 ========
10
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.
15
16 You can use common commands, such as cp and scp, to copy the
17 memory image to a dump file on the local disk, or across the network to
18 a remote system.
19
20 Kdump and kexec are currently supported on the x86, x86_64, ppc64 and ia64
21 architectures.
22
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
27 memory.
28
29 On x86 machines, the first 640 KB of physical memory is needed to boot,
30 regardless of where the kernel loads. Therefore, kexec backs up this
31 region just before rebooting into the dump-capture kernel.
32
33 Similarly on PPC64 machines first 32KB of physical memory is needed for
34 booting regardless of where the kernel is loaded and to support 64K page
35 size kexec backs up the first 64KB memory.
36
37 All of the necessary information about the system kernel's core image is
38 encoded in the ELF format, and stored in a reserved area of memory
39 before a crash. The physical address of the start of the ELF header is
40 passed to the dump-capture kernel through the elfcorehdr= boot
41 parameter.
42
43 With the dump-capture kernel, you can access the memory image, or "old
44 memory," in two ways:
45
46 - Through a /dev/oldmem device interface. A capture utility can read the
47  device file and write out the memory in raw format. This is a raw dump
48  of memory. Analysis and capture tools must be intelligent enough to
49  determine where to look for the right information.
50
51 - Through /proc/vmcore. This exports the dump as an ELF-format file that
52  you can write out using file copy commands such as cp or scp. Further,
53  you can use analysis tools such as the GNU Debugger (GDB) and the Crash
54  tool to debug the dump file. This method ensures that the dump pages are
55  correctly ordered.
56
57
58 Setup and Installation
59 ======================
60
61 Install kexec-tools
62 -------------------
63
64 1) Login as the root user.
65
66 2) Download the kexec-tools user-space package from the following URL:
67
68 http://kernel.org/pub/linux/utils/kernel/kexec/kexec-tools.tar.gz
69
70 This is a symlink to the latest version.
71
72 The latest kexec-tools git tree is available at:
73
74 git://git.kernel.org/pub/scm/utils/kernel/kexec/kexec-tools.git
75 and
76 http://www.kernel.org/pub/scm/utils/kernel/kexec/kexec-tools.git
77
78 There is also a gitweb interface available at
79 http://www.kernel.org/git/?p=utils/kernel/kexec/kexec-tools.git
80
81 More information about kexec-tools can be found at
82 http://www.kernel.org/pub/linux/utils/kernel/kexec/README.html
83
84 3) Unpack the tarball with the tar command, as follows:
85
86   tar xvpzf kexec-tools.tar.gz
87
88 4) Change to the kexec-tools directory, as follows:
89
90   cd kexec-tools-VERSION
91
92 5) Configure the package, as follows:
93
94   ./configure
95
96 6) Compile the package, as follows:
97
98   make
99
100 7) Install the package, as follows:
101
102   make install
103
104
105 Build the system and dump-capture kernels
106 -----------------------------------------
107 There are two possible methods of using Kdump.
108
109 1) Build a separate custom dump-capture kernel for capturing the
110   kernel core dump.
111
112 2) Or use the system kernel binary itself as dump-capture kernel and there is
113   no need to build a separate dump-capture kernel. This is possible
114   only with the architectures which support a relocatable kernel. As
115   of today, i386, x86_64, ppc64 and ia64 architectures support relocatable
116   kernel.
117
118 Building a relocatable kernel is advantageous from the point of view that
119 one does not have to build a second kernel for capturing the dump. But
120 at the same time one might want to build a custom dump capture kernel
121 suitable to his needs.
122
123 Following are the configuration setting required for system and
124 dump-capture kernels for enabling kdump support.
125
126 System kernel config options
127 ----------------------------
128
129 1) Enable "kexec system call" in "Processor type and features."
130
131   CONFIG_KEXEC=y
132
133 2) Enable "sysfs file system support" in "Filesystem" -> "Pseudo
134   filesystems." This is usually enabled by default.
135
136   CONFIG_SYSFS=y
137
138   Note that "sysfs file system support" might not appear in the "Pseudo
139   filesystems" menu if "Configure standard kernel features (for small
140   systems)" is not enabled in "General Setup." In this case, check the
141   .config file itself to ensure that sysfs is turned on, as follows:
142
143   grep 'CONFIG_SYSFS' .config
144
145 3) Enable "Compile the kernel with debug info" in "Kernel hacking."
146
147   CONFIG_DEBUG_INFO=Y
148
149   This causes the kernel to be built with debug symbols. The dump
150   analysis tools require a vmlinux with debug symbols in order to read
151   and analyze a dump file.
152
153 Dump-capture kernel config options (Arch Independent)
154 -----------------------------------------------------
155
156 1) Enable "kernel crash dumps" support under "Processor type and
157   features":
158
159   CONFIG_CRASH_DUMP=y
160
161 2) Enable "/proc/vmcore support" under "Filesystems" -> "Pseudo filesystems".
162
163   CONFIG_PROC_VMCORE=y
164   (CONFIG_PROC_VMCORE is set by default when CONFIG_CRASH_DUMP is selected.)
165
166 Dump-capture kernel config options (Arch Dependent, i386 and x86_64)
167 --------------------------------------------------------------------
168
169 1) On i386, enable high memory support under "Processor type and
170   features":
171
172   CONFIG_HIGHMEM64G=y
173   or
174   CONFIG_HIGHMEM4G
175
176 2) On i386 and x86_64, disable symmetric multi-processing support
177   under "Processor type and features":
178
179   CONFIG_SMP=n
180
181   (If CONFIG_SMP=y, then specify maxcpus=1 on the kernel command line
182   when loading the dump-capture kernel, see section "Load the Dump-capture
183   Kernel".)
184
185 3) If one wants to build and use a relocatable kernel,
186   Enable "Build a relocatable kernel" support under "Processor type and
187   features"
188
189   CONFIG_RELOCATABLE=y
190
191 4) Use a suitable value for "Physical address where the kernel is
192   loaded" (under "Processor type and features"). This only appears when
193   "kernel crash dumps" is enabled. A suitable value depends upon
194   whether kernel is relocatable or not.
195
196   If you are using a relocatable kernel use CONFIG_PHYSICAL_START=0x100000
197   This will compile the kernel for physical address 1MB, but given the fact
198   kernel is relocatable, it can be run from any physical address hence
199   kexec boot loader will load it in memory region reserved for dump-capture
200   kernel.
201
202   Otherwise it should be the start of memory region reserved for
203   second kernel using boot parameter "crashkernel=Y@X". Here X is
204   start of memory region reserved for dump-capture kernel.
205   Generally X is 16MB (0x1000000). So you can set
206   CONFIG_PHYSICAL_START=0x1000000
207
208 5) Make and install the kernel and its modules. DO NOT add this kernel
209   to the boot loader configuration files.
210
211 Dump-capture kernel config options (Arch Dependent, ppc64)
212 ----------------------------------------------------------
213
214 1) Enable "Build a kdump crash kernel" support under "Kernel" options:
215
216   CONFIG_CRASH_DUMP=y
217
218 2)   Enable "Build a relocatable kernel" support
219
220   CONFIG_RELOCATABLE=y
221
222   Make and install the kernel and its modules.
223
224 Dump-capture kernel config options (Arch Dependent, ia64)
225 ----------------------------------------------------------
226
227 - No specific options are required to create a dump-capture kernel
228  for ia64, other than those specified in the arch independent section
229  above. This means that it is possible to use the system kernel
230  as a dump-capture kernel if desired.
231
232  The crashkernel region can be automatically placed by the system
233  kernel at run time. This is done by specifying the base address as 0,
234  or omitting it all together.
235
236  crashkernel=256M@0
237  or
238  crashkernel=256M
239
240  If the start address is specified, note that the start address of the
241  kernel will be aligned to 64Mb, so if the start address is not then
242  any space below the alignment point will be wasted.
243
244
245 Extended crashkernel syntax
246 ===========================
247
248 While the "crashkernel=size[@offset]" syntax is sufficient for most
249 configurations, sometimes it's handy to have the reserved memory dependent
250 on the value of System RAM -- that's mostly for distributors that pre-setup
251 the kernel command line to avoid a unbootable system after some memory has
252 been removed from the machine.
253
254 The syntax is:
255
256    crashkernel=<range1>:<size1>[,<range2>:<size2>,...][@offset]
257    range=start-[end]
258
259    'start' is inclusive and 'end' is exclusive.
260
261 For example:
262
263    crashkernel=512M-2G:64M,2G-:128M
264
265 This would mean:
266
267    1) if the RAM is smaller than 512M, then don't reserve anything
268       (this is the "rescue" case)
269    2) if the RAM size is between 512M and 2G (exclusive), then reserve 64M
270    3) if the RAM size is larger than 2G, then reserve 128M
271
272
273
274 Boot into System Kernel
275 =======================
276
277 1) Update the boot loader (such as grub, yaboot, or lilo) configuration
278   files as necessary.
279
280 2) Boot the system kernel with the boot parameter "crashkernel=Y@X",
281   where Y specifies how much memory to reserve for the dump-capture kernel
282   and X specifies the beginning of this reserved memory. For example,
283   "crashkernel=64M@16M" tells the system kernel to reserve 64 MB of memory
284   starting at physical address 0x01000000 (16MB) for the dump-capture kernel.
285
286   On x86 and x86_64, use "crashkernel=64M@16M".
287
288   On ppc64, use "crashkernel=128M@32M".
289
290   On ia64, 256M@256M is a generous value that typically works.
291   The region may be automatically placed on ia64, see the
292   dump-capture kernel config option notes above.
293
294 Load the Dump-capture Kernel
295 ============================
296
297 After booting to the system kernel, dump-capture kernel needs to be
298 loaded.
299
300 Based on the architecture and type of image (relocatable or not), one
301 can choose to load the uncompressed vmlinux or compressed bzImage/vmlinuz
302 of dump-capture kernel. Following is the summary.
303
304 For i386 and x86_64:
305 - Use vmlinux if kernel is not relocatable.
306 - Use bzImage/vmlinuz if kernel is relocatable.
307 For ppc64:
308 - Use vmlinux
309 For ia64:
310 - Use vmlinux or vmlinuz.gz
311
312
313 If you are using a uncompressed vmlinux image then use following command
314 to load dump-capture kernel.
315
316   kexec -p <dump-capture-kernel-vmlinux-image> \
317   --initrd=<initrd-for-dump-capture-kernel> --args-linux \
318   --append="root=<root-dev> <arch-specific-options>"
319
320 If you are using a compressed bzImage/vmlinuz, then use following command
321 to load dump-capture kernel.
322
323   kexec -p <dump-capture-kernel-bzImage> \
324   --initrd=<initrd-for-dump-capture-kernel> \
325   --append="root=<root-dev> <arch-specific-options>"
326
327 Please note, that --args-linux does not need to be specified for ia64.
328 It is planned to make this a no-op on that architecture, but for now
329 it should be omitted
330
331 Following are the arch specific command line options to be used while
332 loading dump-capture kernel.
333
334 For i386, x86_64 and ia64:
335 "1 irqpoll maxcpus=1 reset_devices"
336
337 For ppc64:
338 "1 maxcpus=1 noirqdistrib reset_devices"
339
340
341 Notes on loading the dump-capture kernel:
342
343 * By default, the ELF headers are stored in ELF64 format to support
344  systems with more than 4GB memory. On i386, kexec automatically checks if
345  the physical RAM size exceeds the 4 GB limit and if not, uses ELF32.
346  So, on non-PAE systems, ELF32 is always used.
347
348  The --elf32-core-headers option can be used to force the generation of ELF32
349  headers. This is necessary because GDB currently cannot open vmcore files
350  with ELF64 headers on 32-bit systems.
351
352 * The "irqpoll" boot parameter reduces driver initialization failures
353  due to shared interrupts in the dump-capture kernel.
354
355 * You must specify <root-dev> in the format corresponding to the root
356  device name in the output of mount command.
357
358 * Boot parameter "1" boots the dump-capture kernel into single-user
359  mode without networking. If you want networking, use "3".
360
361 * We generally don' have to bring up a SMP kernel just to capture the
362  dump. Hence generally it is useful either to build a UP dump-capture
363  kernel or specify maxcpus=1 option while loading dump-capture kernel.
364
365 Kernel Panic
366 ============
367
368 After successfully loading the dump-capture kernel as previously
369 described, the system will reboot into the dump-capture kernel if a
370 system crash is triggered.  Trigger points are located in panic(),
371 die(), die_nmi() and in the sysrq handler (ALT-SysRq-c).
372
373 The following conditions will execute a crash trigger point:
374
375 If a hard lockup is detected and "NMI watchdog" is configured, the system
376 will boot into the dump-capture kernel ( die_nmi() ).
377
378 If die() is called, and it happens to be a thread with pid 0 or 1, or die()
379 is called inside interrupt context or die() is called and panic_on_oops is set,
380 the system will boot into the dump-capture kernel.
381
382 On powerpc systems when a soft-reset is generated, die() is called by all cpus
383 and the system will boot into the dump-capture kernel.
384
385 For testing purposes, you can trigger a crash by using "ALT-SysRq-c",
386 "echo c > /proc/sysrq-trigger" or write a module to force the panic.
387
388 Write Out the Dump File
389 =======================
390
391 After the dump-capture kernel is booted, write out the dump file with
392 the following command:
393
394   cp /proc/vmcore <dump-file>
395
396 You can also access dumped memory as a /dev/oldmem device for a linear
397 and raw view. To create the device, use the following command:
398
399    mknod /dev/oldmem c 1 12
400
401 Use the dd command with suitable options for count, bs, and skip to
402 access specific portions of the dump.
403
404 To see the entire memory, use the following command:
405
406   dd if=/dev/oldmem of=oldmem.001
407
408
409 Analysis
410 ========
411
412 Before analyzing the dump image, you should reboot into a stable kernel.
413
414 You can do limited analysis using GDB on the dump file copied out of
415 /proc/vmcore. Use the debug vmlinux built with -g and run the following
416 command:
417
418   gdb vmlinux <dump-file>
419
420 Stack trace for the task on processor 0, register display, and memory
421 display work fine.
422
423 Note: GDB cannot analyze core files generated in ELF64 format for x86.
424 On systems with a maximum of 4GB of memory, you can generate
425 ELF32-format headers using the --elf32-core-headers kernel option on the
426 dump kernel.
427
428 You can also use the Crash utility to analyze dump files in Kdump
429 format. Crash is available on Dave Anderson's site at the following URL:
430
431   http://people.redhat.com/~anderson/
432
433
434 To Do
435 =====
436
437 1) Provide relocatable kernels for all architectures to help in maintaining
438   multiple kernels for crash_dump, and the same kernel as the system kernel
439   can be used to capture the dump.
440
441
442 Contact
443 =======
444
445 Vivek Goyal (vgoyal@redhat.com)
446 Maneesh Soni (maneesh@in.ibm.com)