1CORE(5) Linux Programmer's Manual CORE(5)
2
6 core - core dump file
7
9 The default action of certain signals is to cause a process to termi‐
10 nate and produce a core dump file, a disk file containing an image of
11 the process's memory at the time of termination. This image can be
12 used in a debugger (e.g., gdb(1)) to inspect the state of the program
13 at the time that it terminated. A list of the signals which cause a
14 process to dump core can be found in signal(7).
15 A process can set its soft RLIMIT_CORE resource limit to place an upper
17 limit on the size of the core dump file that will be produced if it
18 receives a "core dump" signal; see getrlimit(2) for details.
19 There are various circumstances in which a core dump file is not pro‐
21 duced:
22 * The process does not have permission to write the core file. (By
24 default, the core file is called core or core.pid, where pid is the
25 ID of the process that dumped core, and is created in the current
26 working directory. See below for details on naming.) Writing the
27 core file fails if the directory in which it is to be created is
28 nonwritable, or if a file with the same name exists and is not
29 writable or is not a regular file (e.g., it is a directory or a sym‐
30 bolic link).
31 * A (writable, regular) file with the same name as would be used for
33 the core dump already exists, but there is more than one hard link
34 to that file.
35 * The filesystem where the core dump file would be created is full; or
37 has run out of inodes; or is mounted read-only; or the user has
38 reached their quota for the filesystem.
39 * The directory in which the core dump file is to be created does not
41 exist.
42 * The RLIMIT_CORE (core file size) or RLIMIT_FSIZE (file size)
44 resource limits for the process are set to zero; see getrlimit(2)
45 and the documentation of the shell's ulimit command (limit in
46 csh(1)).
47 * The binary being executed by the process does not have read permis‐
49 sion enabled.
50 * The process is executing a set-user-ID (set-group-ID) program that
52 is owned by a user (group) other than the real user (group) ID of
53 the process, or the process is executing a program that has file
54 capabilities (see capabilities(7)). (However, see the description
55 of the prctl(2) PR_SET_DUMPABLE operation, and the description of
56 the /proc/sys/fs/suid_dumpable file in proc(5).)
57 * /proc/sys/kernel/core_pattern is empty and /proc/sys/ker‐
59 nel/core_uses_pid contains the value 0. (These files are described
60 below.) Note that if /proc/sys/kernel/core_pattern is empty and
61 /proc/sys/kernel/core_uses_pid contains the value 1, core dump files
62 will have names of the form .pid, and such files are hidden unless
63 one uses the ls(1) -a option.
64 * (Since Linux 3.7) The kernel was configured without the CONFIG_CORE‐
66 DUMP option.
67 In addition, a core dump may exclude part of the address space of the
69 process if the madvise(2) MADV_DONTDUMP flag was employed.
70 On systems that employ systemd(1) as the init framework, core dumps may
72 instead be placed in a location determined by systemd(1). See below
73 for further details.
74 Naming of core dump files
76 By default, a core dump file is named core, but the /proc/sys/ker‐
77 nel/core_pattern file (since Linux 2.6 and 2.4.21) can be set to define
78 a template that is used to name core dump files. The template can con‐
79 tain % specifiers which are substituted by the following values when a
80 core file is created:
81 %% a single % character
83 %c core file size soft resource limit of crashing process (since
84 Linux 2.6.24)
85 %d dump mode—same as value returned by prctl(2) PR_GET_DUMPABLE
86 (since Linux 3.7)
87 %e executable filename (without path prefix)
88 %E pathname of executable, with slashes ('/') replaced by exclama‐
89 tion marks ('!') (since Linux 3.0).
90 %g (numeric) real GID of dumped process
91 %h hostname (same as nodename returned by uname(2))
92 %i TID of thread that triggered core dump, as seen in the PID
93 namespace in which the thread resides (since Linux 3.18)
94 %I TID of thread that triggered core dump, as seen in the initial
95 PID namespace (since Linux 3.18)
96 %p PID of dumped process, as seen in the PID namespace in which
97 the process resides
98 %P PID of dumped process, as seen in the initial PID namespace
99 (since Linux 3.12)
100 %s number of signal causing dump
101 %t time of dump, expressed as seconds since the Epoch, 1970-01-01
102 00:00:00 +0000 (UTC)
103 %u (numeric) real UID of dumped process
104 A single % at the end of the template is dropped from the core file‐
106 name, as is the combination of a % followed by any character other than
107 those listed above. All other characters in the template become a lit‐
108 eral part of the core filename. The template may include '/' charac‐
109 ters, which are interpreted as delimiters for directory names. The
110 maximum size of the resulting core filename is 128 bytes (64 bytes in
111 kernels before 2.6.19). The default value in this file is "core". For
112 backward compatibility, if /proc/sys/kernel/core_pattern does not
113 include %p and /proc/sys/kernel/core_uses_pid (see below) is nonzero,
114 then .PID will be appended to the core filename.
115 Paths are interpreted according to the settings that are active for the
117 crashing process. That means the crashing process's mount namespace
118 (see mount_namespaces(7)), its current working directory (found via
119 getcwd(2)), and its root directory (see chroot(2)).
120 Since version 2.4, Linux has also provided a more primitive method of
122 controlling the name of the core dump file. If the /proc/sys/ker‐
123 nel/core_uses_pid file contains the value 0, then a core dump file is
124 simply named core. If this file contains a nonzero value, then the
125 core dump file includes the process ID in a name of the form core.PID.
126 Since Linux 3.6, if /proc/sys/fs/suid_dumpable is set to 2 ("suid‐
128 safe"), the pattern must be either an absolute pathname (starting with
129 a leading '/' character) or a pipe, as defined below.
130 Piping core dumps to a program
132 Since kernel 2.6.19, Linux supports an alternate syntax for the
133 /proc/sys/kernel/core_pattern file. If the first character of this
134 file is a pipe symbol (|), then the remainder of the line is inter‐
135 preted as the command-line for a user-space program (or script) that is
136 to be executed.
137 Since kernel 5.3.0, the pipe template is split on spaces into an argu‐
139 ment list before the template parameters are expanded. In earlier ker‐
140 nels, the template parameters are expanded first and the resulting
141 string is split on spaces into an argument list. This means that in
142 earlier kernels executable names added by the %e and %E template param‐
143 eters could get split into multiple arguments. So the core dump han‐
144 dler needs to put the executable names as the last argument and ensure
145 it joins all parts of the executable name using spaces. Executable
146 names with multiple spaces in them are not correctly represented in
147 earlier kernels, meaning that the core dump handler needs to use mecha‐
148 nisms to find the executable name.
149 Instead of being written to a disk file, the core dump is given as
151 standard input to the program. Note the following points:
152 * The program must be specified using an absolute pathname (or a path‐
154 name relative to the root directory, /), and must immediately follow
155 the '|' character.
156 * The command-line arguments can include any of the % specifiers
158 listed above. For example, to pass the PID of the process that is
159 being dumped, specify %p in an argument.
160 * The process created to run the program runs as user and group root.
162 * Running as root does not confer any exceptional security bypasses.
164 Namely, LSMs (e.g., SELinux) are still active and may prevent the
165 handler from accessing details about the crashed process via
166 /proc/[pid].
167 * The program pathname is interpreted with respect to the initial
169 mount namespace as it is always executed there. It is not affected
170 by the settings (e.g., root directory, mount namespace, current
171 working directory) of the crashing process.
172 * The process runs in the initial namespaces (PID, mount, user, and so
174 on) and not in the namespaces of the crashing process. One can uti‐
175 lize specifiers such as %P to find the right /proc/[pid] directory
176 and probe/enter the crashing process's namespaces if needed.
177 * The process starts with its current working directory as the root
179 directory. If desired, it is possible change to the working direc‐
180 tory of the dumping process by employing the value provided by the
181 %P specifier to change to the location of the dumping process via
182 /proc/[pid]/cwd.
183 * Command-line arguments can be supplied to the program (since Linux
185 2.6.24), delimited by white space (up to a total line length of 128
186 bytes).
187 * The RLIMIT_CORE limit is not enforced for core dumps that are piped
189 to a program via this mechanism.
190 /proc/sys/kernel/core_pipe_limit
192 When collecting core dumps via a pipe to a user-space program, it can
193 be useful for the collecting program to gather data about the crashing
194 process from that process's /proc/[pid] directory. In order to do this
195 safely, the kernel must wait for the program collecting the core dump
196 to exit, so as not to remove the crashing process's /proc/[pid] files
197 prematurely. This in turn creates the possibility that a misbehaving
198 collecting program can block the reaping of a crashed process by simply
199 never exiting.
200 Since Linux 2.6.32, the /proc/sys/kernel/core_pipe_limit can be used to
202 defend against this possibility. The value in this file defines how
203 many concurrent crashing processes may be piped to user-space programs
204 in parallel. If this value is exceeded, then those crashing processes
205 above this value are noted in the kernel log and their core dumps are
206 skipped.
207 A value of 0 in this file is special. It indicates that unlimited pro‐
209 cesses may be captured in parallel, but that no waiting will take place
210 (i.e., the collecting program is not guaranteed access to /proc/<crash‐
211 ing-PID>). The default value for this file is 0.
212 Controlling which mappings are written to the core dump
214 Since kernel 2.6.23, the Linux-specific /proc/[pid]/coredump_filter
215 file can be used to control which memory segments are written to the
216 core dump file in the event that a core dump is performed for the
217 process with the corresponding process ID.
218 The value in the file is a bit mask of memory mapping types (see
220 mmap(2)). If a bit is set in the mask, then memory mappings of the
221 corresponding type are dumped; otherwise they are not dumped. The bits
222 in this file have the following meanings:
223 bit 0 Dump anonymous private mappings.
225 bit 1 Dump anonymous shared mappings.
226 bit 2 Dump file-backed private mappings.
227 bit 3 Dump file-backed shared mappings.
228 bit 4 (since Linux 2.6.24)
229 Dump ELF headers.
230 bit 5 (since Linux 2.6.28)
231 Dump private huge pages.
232 bit 6 (since Linux 2.6.28)
233 Dump shared huge pages.
234 bit 7 (since Linux 4.4)
235 Dump private DAX pages.
236 bit 8 (since Linux 4.4)
237 Dump shared DAX pages.
238 By default, the following bits are set: 0, 1, 4 (if the CON‐
240 FIG_CORE_DUMP_DEFAULT_ELF_HEADERS kernel configuration option is
241 enabled), and 5. This default can be modified at boot time using the
242 coredump_filter boot option.
243 The value of this file is displayed in hexadecimal. (The default value
245 is thus displayed as 33.)
246 Memory-mapped I/O pages such as frame buffer are never dumped, and vir‐
248 tual DSO (vdso(7)) pages are always dumped, regardless of the core‐
249 dump_filter value.
250 A child process created via fork(2) inherits its parent's coredump_fil‐
252 ter value; the coredump_filter value is preserved across an execve(2).
253 It can be useful to set coredump_filter in the parent shell before run‐
255 ning a program, for example:
256 $ echo 0x7 > /proc/self/coredump_filter
258 $ ./some_program
259 This file is provided only if the kernel was built with the CON‐
261 FIG_ELF_CORE configuration option.
262 Core dumps and systemd
264 On systems using the systemd(1) init framework, core dumps may be
265 placed in a location determined by systemd(1). To do this, systemd(1)
266 employs the core_pattern feature that allows piping core dumps to a
267 program. One can verify this by checking whether core dumps are being
268 piped to the systemd-coredump(8) program:
269 $ cat /proc/sys/kernel/core_pattern
271 |/usr/lib/systemd/systemd-coredump %P %u %g %s %t %c %e
272 In this case, core dumps will be placed in the location configured for
274 systemd-coredump(8), typically as lz4(1) compressed files in the direc‐
275 tory /var/lib/systemd/coredump/. One can list the core dumps that have
276 been recorded by systemd-coredump(8) using coredumpctl(1):
277 $ coredumpctl list | tail -5
279 Wed 2017-10-11 22:25:30 CEST 2748 1000 1000 3 present /usr/bin/sleep
280 Thu 2017-10-12 06:29:10 CEST 2716 1000 1000 3 present /usr/bin/sleep
281 Thu 2017-10-12 06:30:50 CEST 2767 1000 1000 3 present /usr/bin/sleep
282 Thu 2017-10-12 06:37:40 CEST 2918 1000 1000 3 present /usr/bin/cat
283 Thu 2017-10-12 08:13:07 CEST 2955 1000 1000 3 present /usr/bin/cat
284 The information shown for each core dump includes the date and time of
286 the dump, the PID, UID, and GID of the dumping process, the signal
287 number that caused the core dump, and the pathname of the executable
288 that was being run by the dumped process. Various options to core‐
289 dumpctl(1) allow a specified coredump file to be pulled from the sys‐
290 temd(1) location into a specified file. For example, to extract the
291 core dump for PID 2955 shown above to a file named core in the current
292 directory, one could use:
293 $ coredumpctl dump 2955 -o core
295 For more extensive details, see the coredumpctl(1) manual page.
297 To disable the systemd(1) mechanism that archives core dumps, restoring
299 to something more like traditional Linux behavior, one can set an over‐
300 ride for the systemd(1) mechanism, using something like:
301 # echo "kernel.core_pattern=core.%p" > /etc/sysctl.d/50-coredump.conf
303 # /lib/systemd/systemd-sysctl
304
306 The gdb(1) gcore command can be used to obtain a core dump of a running
307 process.
308 In Linux versions up to and including 2.6.27, if a multithreaded
310 process (or, more precisely, a process that shares its memory with
311 another process by being created with the CLONE_VM flag of clone(2))
312 dumps core, then the process ID is always appended to the core file‐
313 name, unless the process ID was already included elsewhere in the file‐
314 name via a %p specification in /proc/sys/kernel/core_pattern. (This is
315 primarily useful when employing the obsolete LinuxThreads implementa‐
316 tion, where each thread of a process has a different PID.)
317
319 The program below can be used to demonstrate the use of the pipe syntax
320 in the /proc/sys/kernel/core_pattern file. The following shell session
321 demonstrates the use of this program (compiled to create an executable
322 named core_pattern_pipe_test):
323 $ cc -o core_pattern_pipe_test core_pattern_pipe_test.c
325 $ su
326 Password:
327 # echo "|$PWD/core_pattern_pipe_test %p UID=%u GID=%g sig=%s" > \
328 /proc/sys/kernel/core_pattern
329 # exit
330 $ sleep 100
331 ^\ # type control-backslash
332 Quit (core dumped)
333 $ cat core.info
334 argc=5
335 argc[0]=</home/mtk/core_pattern_pipe_test>
336 argc[1]=<20575>
337 argc[2]=<UID=1000>
338 argc[3]=<GID=100>
339 argc[4]=<sig=3>
340 Total bytes in core dump: 282624
341 Program source
343 /* core_pattern_pipe_test.c */
345 #define _GNU_SOURCE
347 #include <sys/stat.h>
348 #include <fcntl.h>
349 #include <limits.h>
350 #include <stdio.h>
351 #include <stdlib.h>
352 #include <unistd.h>
353 #define BUF_SIZE 1024
355 int
357 main(int argc, char *argv[])
358 {
359 int tot, j;
360 ssize_t nread;
361 char buf[BUF_SIZE];
362 FILE *fp;
363 char cwd[PATH_MAX];
364 /* Change our current working directory to that of the
366 crashing process */
367 snprintf(cwd, PATH_MAX, "/proc/%s/cwd", argv[1]);
369 chdir(cwd);
370 /* Write output to file "core.info" in that directory */
372 fp = fopen("core.info", "w+");
374 if (fp == NULL)
375 exit(EXIT_FAILURE);
376 /* Display command-line arguments given to core_pattern
378 pipe program */
379 fprintf(fp, "argc=%d\n", argc);
381 for (j = 0; j < argc; j++)
382 fprintf(fp, "argc[%d]=<%s>\n", j, argv[j]);
383 /* Count bytes in standard input (the core dump) */
385 tot = 0;
387 while ((nread = read(STDIN_FILENO, buf, BUF_SIZE)) > 0)
388 tot += nread;
389 fprintf(fp, "Total bytes in core dump: %d\n", tot);
390 fclose(fp);
392 exit(EXIT_SUCCESS);
393 }
394
396 bash(1), coredumpctl(1), gdb(1), getrlimit(2), mmap(2), prctl(2),
397 sigaction(2), elf(5), proc(5), pthreads(7), signal(7), systemd-core‐
398 dump(8)
399
401 This page is part of release 5.04 of the Linux man-pages project. A
402 description of the project, information about reporting bugs, and the
403 latest version of this page, can be found at
404 https://www.kernel.org/doc/man-pages/.
405Linux 2019-10-10 CORE(5)