-
Notifications
You must be signed in to change notification settings - Fork 278
Linux kernel crash utility
crash-utility/crash
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
CORE ANALYSIS SUITE
The core analysis suite is a self-contained tool that can be used to
investigate either live systems, kernel core dumps created from dump
creation facilities such as kdump, kvmdump, xendump, the netdump and
diskdump packages offered by Red Hat, the LKCD kernel patch, the mcore
kernel patch created by Mission Critical Linux, as well as other formats
created by manufacturer-specific firmware.
o The tool is loosely based on the SVR4 crash command, but has been
completely integrated with gdb in order to be able to display
formatted kernel data structures, disassemble source code, etc.
o The current set of available commands consist of common kernel core
analysis tools such as a context-specific stack traces, source code
disassembly, kernel variable displays, memory display, dumps of
linked-lists, etc. In addition, any gdb command may be entered,
which in turn will be passed onto the gdb module for execution.
o There are several commands that delve deeper into specific kernel
subsystems, which also serve as templates for kernel developers
to create new commands for analysis of a specific area of interest.
Adding a new command is a simple affair, and a quick recompile
adds it to the command menu.
o The intent is to make the tool independent of Linux version dependencies,
building in recognition of major kernel code changes so as to adapt to
new kernel versions, while maintaining backwards compatibility.
A whitepaper with complete documentation concerning the use of this utility
can be found here:
https://crash-utility.github.io/crash_whitepaper.html
These are the current prerequisites:
o At this point, x86, ia64, x86_64, ppc64, ppc, arm, arm64, alpha, mips,
mips64, loongarch64, riscv64, s390 and s390x-based kernels are supported.
Other architectures may be addressed in the future.
o One size fits all -- the utility can be run on any Linux kernel version
version dating back to 2.2.5-15. A primary design goal is to always
maintain backwards-compatibility.
o In order to contain debugging data, the top-level kernel Makefile's CFLAGS
definition must contain the -g flag. Typically distributions will contain
a package containing a vmlinux file with full debuginfo data. If not, the
kernel must be rebuilt:
For 2.2 kernels that are not built with -g, change the following line:
CFLAGS = -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer
to:
CFLAGS = -g -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer
For 2.4 kernels that are not built with -g, change the following line:
CFLAGS := $(CPPFLAGS) -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer -fno-strict-aliasing
to:
CFLAGS := -g $(CPPFLAGS) -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer -fno-strict-aliasing
For 2.6 and later kernels that are not built with -g, the kernel should
be configured with CONFIG_DEBUG_INFO enabled, which in turn will add
the -g flag to the CFLAGS setting in the kernel Makefile.
After the kernel is re-compiled, the uncompressed "vmlinux" kernel
that is created in the top-level kernel build directory must be saved.
To build the crash utility:
$ tar -xf crash-8.0.6.tar.gz
$ cd crash-8.0.6
$ make
The initial build will take several minutes because the embedded gdb module
must be configured and built. Alternatively, the crash source RPM file
may be installed and built, and the resultant crash binary RPM file installed.
The crash binary can only be used on systems of the same architecture as
the host build system. There are a few optional manners of building the
crash binary:
o On an x86_64 host, a 32-bit x86 binary that can be used to analyze
32-bit x86 dumpfiles may be built by typing "make target=X86".
o On an x86 or x86_64 host, a 32-bit x86 binary that can be used to analyze
32-bit arm dumpfiles may be built by typing "make target=ARM".
o On an x86 or x86_64 host, a 32-bit x86 binary that can be used to analyze
32-bit mips dumpfiles may be built by typing "make target=MIPS".
o On an ppc64 host, a 32-bit ppc binary that can be used to analyze
32-bit ppc dumpfiles may be built by typing "make target=PPC".
o On an x86_64 host, an x86_64 binary that can be used to analyze
arm64 dumpfiles may be built by typing "make target=ARM64".
o On an x86_64 host, an x86_64 binary that can be used to analyze
ppc64le dumpfiles may be built by typing "make target=PPC64".
o On an x86_64 host, an x86_64 binary that can be used to analyze
riscv64 dumpfiles may be built by typing "make target=RISCV64".
o On an x86_64 host, an x86_64 binary that can be used to analyze
loongarch64 dumpfiles may be built by typing "make target=LOONGARCH64".
Traditionally when vmcores are compressed via the makedumpfile(8) facility
the libz compression library is used, and by default the crash utility
only supports libz. Recently makedumpfile has been enhanced to optionally
use the LZO, snappy or zstd compression libraries. To build crash with any
or all of those libraries, type "make lzo", "make snappy" or "make zstd".
crash supports valgrind Memcheck tool on the crash's custom memory allocator.
To build crash with this feature enabled, type "make valgrind" and then run
crash with valgrind as "valgrind crash vmlinux vmcore".
All of the alternate build commands above are "sticky" in that the
special "make" targets only have to be entered one time; all subsequent
builds will follow suit.
If the tool is run against a kernel dumpfile, two arguments are required, the
uncompressed kernel name and the kernel dumpfile name.
If run on a live system, only the kernel name is required, because /dev/mem
will be used as the "dumpfile". On Red Hat or Fedora kernels where the
/dev/mem device is restricted, the /dev/crash memory driver will be used.
If neither /dev/mem or /dev/crash are available, then /proc/kcore will be
be used as the live memory source. If /proc/kcore is also restricted, then
the Red Hat /dev/crash driver may be compiled and installed; its source
is included in the crash-8.0.6/memory_driver subdirectory.
If the kernel file is stored in /boot, /, /boot/efi, or in any /usr/src
or /usr/lib/debug/lib/modules subdirectory, then no command line arguments
are required -- the first kernel found that matches /proc/version will be
used as the namelist.
For example, invoking crash on a live system would look like this:
$ crash
crash 8.0.6
Copyright (C) 2002-2022 Red Hat, Inc.
Copyright (C) 2004, 2005, 2006, 2010 IBM Corporation
Copyright (C) 1999-2006 Hewlett-Packard Co
Copyright (C) 2005, 2006, 2011, 2012 Fujitsu Limited
Copyright (C) 2006, 2007 VA Linux Systems Japan K.K.
Copyright (C) 2005, 2011, 2020-2022 NEC Corporation
Copyright (C) 1999, 2002, 2007 Silicon Graphics, Inc.
Copyright (C) 1999, 2000, 2001, 2002 Mission Critical Linux, Inc.
Copyright (C) 2015, 2021 VMware, Inc.
This program is free software, covered by the GNU General Public License,
and you are welcome to change it and/or distribute copies of it under
certain conditions. Enter "help copying" to see the conditions.
This program has absolutely no warranty. Enter "help warranty" for details.
GNU gdb 10.2
Copyright 2013 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "i686-pc-linux-gnu"...
KERNEL: /boot/vmlinux
DUMPFILE: /dev/mem
CPUS: 1
DATE: Tue Nov 12 09:25:55 CST 2024
UPTIME: 10 days, 22:55:18
LOAD AVERAGE: 0.08, 0.03, 0.01
TASKS: 42
NODENAME: ha2.mclinux.com
RELEASE: 2.4.0-test10
VERSION: #11 SMP Thu Nov 4 15:09:25 EST 2000
MACHINE: i686 (447 MHz)
MEMORY: 128 MB
PID: 3621
COMMAND: "crash"
TASK: c463c000
CPU: 0
STATE: TASK_RUNNING (ACTIVE)
crash> help
* files mod sbitmapq union
alias foreach mount search vm
ascii fuser net set vtop
bpf gdb p sig waitq
bt help ps struct whatis
btop ipcs pte swap wr
dev irq ptob sym q
dis kmem ptov sys
eval list rd task
exit log repeat timer
extend mach runq tree
crash version: 8.0.6 gdb version: 10.2
For help on any command above, enter "help <command>".
For help on input options, enter "help input".
For help on output options, enter "help output".
crash>
When run on a dumpfile, both the kernel namelist and dumpfile must be
entered on the command line. For example, when run on a core dump created
by the Red Hat netdump or diskdump facilities:
$ crash vmlinux vmcore
crash 8.0.6
Copyright (C) 2002-2022 Red Hat, Inc.
Copyright (C) 2004, 2005, 2006, 2010 IBM Corporation
Copyright (C) 1999-2006 Hewlett-Packard Co
Copyright (C) 2005, 2006, 2011, 2012 Fujitsu Limited
Copyright (C) 2006, 2007 VA Linux Systems Japan K.K.
Copyright (C) 2005, 2011, 2020-2022 NEC Corporation
Copyright (C) 1999, 2002, 2007 Silicon Graphics, Inc.
Copyright (C) 1999, 2000, 2001, 2002 Mission Critical Linux, Inc.
Copyright (C) 2015, 2021 VMware, Inc.
This program is free software, covered by the GNU General Public License,
and you are welcome to change it and/or distribute copies of it under
certain conditions. Enter "help copying" to see the conditions.
This program has absolutely no warranty. Enter "help warranty" for details.
GNU gdb 10.2
Copyright 2013 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "i686-pc-linux-gnu"...
KERNEL: vmlinux
DUMPFILE: vmcore
CPUS: 4
DATE: Tue Mar 2 13:57:09 2004
UPTIME: 00:02:40
LOAD AVERAGE: 2.24, 0.96, 0.37
TASKS: 70
NODENAME: pro1.lab.boston.redhat.com
RELEASE: 2.6.3-2.1.214.11smp
VERSION: #1 SMP Tue Mar 2 10:58:27 EST 2004
MACHINE: i686 (2785 Mhz)
MEMORY: 512 MB
PANIC: "Oops: 0002 [#1]" (check log for details)
PID: 0
COMMAND: "swapper"
TASK: 22fa200 (1 of 4) [THREAD_INFO: 2356000]
CPU: 0
STATE: TASK_RUNNING (PANIC)
crash>
The tool's environment is context-specific. On a live system, the default
context is the command itself; on a dump the default context will be the
task that panicked. The most commonly-used commands are:
set - set a new task context by pid, task address, or cpu.
bt - backtrace of the current context, or as specified with arguments.
p - print the contents of a kernel variable.
rd - read memory, which may be either kernel virtual, user virtual, or
physical.
ps - simple process listing.
log - dump the kernel log_buf.
struct - print the contents of a structure at a specified address.
foreach - execute a command on all tasks, or those specified, in the system.
Detailed help concerning the use of each of the commands in the menu above
may be displayed by entering "help command", where "command" is one of those
listed above. Rather than getting bogged down in details here, simply
run the help command on each of the commands above. Note that many commands
have multiple options so as to avoid the proliferation of command names.
Command output may be piped to external commands or redirected to files.
Enter "help output" for details.
The command line history mechanism allows for command-line recall and
command-line editing. Input files containing a set of crash commands may
be substituted for command-line input. Enter "help input" for details.
Note that a .crashrc file (or .<your-command-name>rc if the name has been
changed), may contain any number of "set" or "alias" commands -- see the
help pages on those two commands for details.
Lastly, if a command is entered that is not recognized, it is checked
against the kernel's list of variables, structure, union or typedef names,
and if found, the command is passed to "p", "struct", "union" or "whatis".
That being the case, as long as a kernel variable/structure/union name is
different than any of the current commands.
(1) A kernel variable can be dumped by simply entering its name:
crash> init_mm
init_mm = $2 = {
mmap = 0xc022d540,
mmap_avl = 0x0,
mmap_cache = 0x0,
pgd = 0xc0101000,
count = {
counter = 0x6
},
map_count = 0x1,
mmap_sem = {
count = {
counter = 0x1
},
waking = 0x0,
wait = 0x0
},
context = 0x0,
start_code = 0xc0000000,
end_code = 0xc022b4c8,
end_data = c0250388,
...
(2) A structure or can be dumped simply by entering its name and address:
crash> vm_area_struct c5ba3910
struct vm_area_struct {
vm_mm = 0xc3ae3210,
vm_start = 0x821b000,
vm_end = 0x8692000,
vm_next = 0xc5ba3890,
vm_page_prot = {
pgprot = 0x25
},
vm_flags = 0x77,
vm_avl_height = 0x4,
vm_avl_left = 0xc0499540,
vm_avl_right = 0xc0499f40,
vm_next_share = 0xc04993c0,
vm_pprev_share = 0xc0499060,
vm_ops = 0x0,
vm_offset = 0x0,
vm_file = 0x0,
vm_pte = 0x0
}
The crash utility has been designed to facilitate the task of adding new
commands. New commands may be permanently compiled into the crash executable,
or dynamically added during runtime using shared object files.
To permanently add a new command to the crash executable's menu:
1. For a command named "xxx", put a reference to cmd_xxx() in defs.h.
2. Add cmd_xxx into the base_command_table[] array in global_data.c.
3. Write cmd_xxx(), putting it in one of the appropriate files. Look at
the other commands for guidance on getting symbolic data, reading
memory, displaying data, etc...
4. Recompile and run.
Note that while the initial compile of crash, which configures and compiles
the gdb module, takes several minutes, subsequent re-compiles to do such
things as add new commands or fix bugs just takes a few seconds.
Alternatively, you can create shared object library files consisting of
crash command extensions, that can be dynamically linked into the crash
executable during runtime or during initialization. This will allow
the same shared object to be used with subsequent crash releases without
having to re-merge the command's code into each new set of crash sources.
The dynamically linked-in commands will automatically show up in the crash
help menu. For details, enter "help extend" during runtime, or enter
"crash -h extend" from the shell command line.