1getrlimit(2) System Calls Manual getrlimit(2)
2
3
4
6 getrlimit, setrlimit, prlimit - get/set resource limits
7
9 Standard C library (libc, -lc)
10
12 #include <sys/resource.h>
13
14 int getrlimit(int resource, struct rlimit *rlim);
15 int setrlimit(int resource, const struct rlimit *rlim);
16
17 int prlimit(pid_t pid, int resource,
18 const struct rlimit *_Nullable new_limit,
19 struct rlimit *_Nullable old_limit);
20
21 Feature Test Macro Requirements for glibc (see feature_test_macros(7)):
22
23 prlimit():
24 _GNU_SOURCE
25
27 The getrlimit() and setrlimit() system calls get and set resource lim‐
28 its. Each resource has an associated soft and hard limit, as defined
29 by the rlimit structure:
30
31 struct rlimit {
32 rlim_t rlim_cur; /* Soft limit */
33 rlim_t rlim_max; /* Hard limit (ceiling for rlim_cur) */
34 };
35
36 The soft limit is the value that the kernel enforces for the corre‐
37 sponding resource. The hard limit acts as a ceiling for the soft
38 limit: an unprivileged process may set only its soft limit to a value
39 in the range from 0 up to the hard limit, and (irreversibly) lower its
40 hard limit. A privileged process (under Linux: one with the
41 CAP_SYS_RESOURCE capability in the initial user namespace) may make ar‐
42 bitrary changes to either limit value.
43
44 The value RLIM_INFINITY denotes no limit on a resource (both in the
45 structure returned by getrlimit() and in the structure passed to setr‐
46 limit()).
47
48 The resource argument must be one of:
49
50 RLIMIT_AS
51 This is the maximum size of the process's virtual memory (ad‐
52 dress space). The limit is specified in bytes, and is rounded
53 down to the system page size. This limit affects calls to
54 brk(2), mmap(2), and mremap(2), which fail with the error ENOMEM
55 upon exceeding this limit. In addition, automatic stack expan‐
56 sion fails (and generates a SIGSEGV that kills the process if no
57 alternate stack has been made available via sigaltstack(2)).
58 Since the value is a long, on machines with a 32-bit long either
59 this limit is at most 2 GiB, or this resource is unlimited.
60
61 RLIMIT_CORE
62 This is the maximum size of a core file (see core(5)) in bytes
63 that the process may dump. When 0 no core dump files are cre‐
64 ated. When nonzero, larger dumps are truncated to this size.
65
66 RLIMIT_CPU
67 This is a limit, in seconds, on the amount of CPU time that the
68 process can consume. When the process reaches the soft limit,
69 it is sent a SIGXCPU signal. The default action for this signal
70 is to terminate the process. However, the signal can be caught,
71 and the handler can return control to the main program. If the
72 process continues to consume CPU time, it will be sent SIGXCPU
73 once per second until the hard limit is reached, at which time
74 it is sent SIGKILL. (This latter point describes Linux behav‐
75 ior. Implementations vary in how they treat processes which
76 continue to consume CPU time after reaching the soft limit.
77 Portable applications that need to catch this signal should per‐
78 form an orderly termination upon first receipt of SIGXCPU.)
79
80 RLIMIT_DATA
81 This is the maximum size of the process's data segment (initial‐
82 ized data, uninitialized data, and heap). The limit is speci‐
83 fied in bytes, and is rounded down to the system page size.
84 This limit affects calls to brk(2), sbrk(2), and (since Linux
85 4.7) mmap(2), which fail with the error ENOMEM upon encountering
86 the soft limit of this resource.
87
88 RLIMIT_FSIZE
89 This is the maximum size in bytes of files that the process may
90 create. Attempts to extend a file beyond this limit result in
91 delivery of a SIGXFSZ signal. By default, this signal termi‐
92 nates a process, but a process can catch this signal instead, in
93 which case the relevant system call (e.g., write(2), trun‐
94 cate(2)) fails with the error EFBIG.
95
96 RLIMIT_LOCKS (Linux 2.4.0 to Linux 2.4.24)
97 This is a limit on the combined number of flock(2) locks and fc‐
98 ntl(2) leases that this process may establish.
99
100 RLIMIT_MEMLOCK
101 This is the maximum number of bytes of memory that may be locked
102 into RAM. This limit is in effect rounded down to the nearest
103 multiple of the system page size. This limit affects mlock(2),
104 mlockall(2), and the mmap(2) MAP_LOCKED operation. Since Linux
105 2.6.9, it also affects the shmctl(2) SHM_LOCK operation, where
106 it sets a maximum on the total bytes in shared memory segments
107 (see shmget(2)) that may be locked by the real user ID of the
108 calling process. The shmctl(2) SHM_LOCK locks are accounted for
109 separately from the per-process memory locks established by
110 mlock(2), mlockall(2), and mmap(2) MAP_LOCKED; a process can
111 lock bytes up to this limit in each of these two categories.
112
113 Before Linux 2.6.9, this limit controlled the amount of memory
114 that could be locked by a privileged process. Since Linux
115 2.6.9, no limits are placed on the amount of memory that a priv‐
116 ileged process may lock, and this limit instead governs the
117 amount of memory that an unprivileged process may lock.
118
119 RLIMIT_MSGQUEUE (since Linux 2.6.8)
120 This is a limit on the number of bytes that can be allocated for
121 POSIX message queues for the real user ID of the calling
122 process. This limit is enforced for mq_open(3). Each message
123 queue that the user creates counts (until it is removed) against
124 this limit according to the formula:
125
126 Since Linux 3.5:
127
128 bytes = attr.mq_maxmsg * sizeof(struct msg_msg) +
129 MIN(attr.mq_maxmsg, MQ_PRIO_MAX) *
130 sizeof(struct posix_msg_tree_node)+
131 /* For overhead */
132 attr.mq_maxmsg * attr.mq_msgsize;
133 /* For message data */
134
135 Linux 3.4 and earlier:
136
137 bytes = attr.mq_maxmsg * sizeof(struct msg_msg *) +
138 /* For overhead */
139 attr.mq_maxmsg * attr.mq_msgsize;
140 /* For message data */
141
142 where attr is the mq_attr structure specified as the fourth ar‐
143 gument to mq_open(3), and the msg_msg and posix_msg_tree_node
144 structures are kernel-internal structures.
145
146 The "overhead" addend in the formula accounts for overhead bytes
147 required by the implementation and ensures that the user cannot
148 create an unlimited number of zero-length messages (such mes‐
149 sages nevertheless each consume some system memory for bookkeep‐
150 ing overhead).
151
152 RLIMIT_NICE (since Linux 2.6.12, but see BUGS below)
153 This specifies a ceiling to which the process's nice value can
154 be raised using setpriority(2) or nice(2). The actual ceiling
155 for the nice value is calculated as 20 - rlim_cur. The useful
156 range for this limit is thus from 1 (corresponding to a nice
157 value of 19) to 40 (corresponding to a nice value of -20). This
158 unusual choice of range was necessary because negative numbers
159 cannot be specified as resource limit values, since they typi‐
160 cally have special meanings. For example, RLIM_INFINITY typi‐
161 cally is the same as -1. For more detail on the nice value, see
162 sched(7).
163
164 RLIMIT_NOFILE
165 This specifies a value one greater than the maximum file de‐
166 scriptor number that can be opened by this process. Attempts
167 (open(2), pipe(2), dup(2), etc.) to exceed this limit yield the
168 error EMFILE. (Historically, this limit was named RLIMIT_OFILE
169 on BSD.)
170
171 Since Linux 4.5, this limit also defines the maximum number of
172 file descriptors that an unprivileged process (one without the
173 CAP_SYS_RESOURCE capability) may have "in flight" to other pro‐
174 cesses, by being passed across UNIX domain sockets. This limit
175 applies to the sendmsg(2) system call. For further details, see
176 unix(7).
177
178 RLIMIT_NPROC
179 This is a limit on the number of extant process (or, more pre‐
180 cisely on Linux, threads) for the real user ID of the calling
181 process. So long as the current number of processes belonging
182 to this process's real user ID is greater than or equal to this
183 limit, fork(2) fails with the error EAGAIN.
184
185 The RLIMIT_NPROC limit is not enforced for processes that have
186 either the CAP_SYS_ADMIN or the CAP_SYS_RESOURCE capability, or
187 run with real user ID 0.
188
189 RLIMIT_RSS
190 This is a limit (in bytes) on the process's resident set (the
191 number of virtual pages resident in RAM). This limit has effect
192 only in Linux 2.4.x, x < 30, and there affects only calls to
193 madvise(2) specifying MADV_WILLNEED.
194
195 RLIMIT_RTPRIO (since Linux 2.6.12, but see BUGS)
196 This specifies a ceiling on the real-time priority that may be
197 set for this process using sched_setscheduler(2) and sched_set‐
198 param(2).
199
200 For further details on real-time scheduling policies, see
201 sched(7)
202
203 RLIMIT_RTTIME (since Linux 2.6.25)
204 This is a limit (in microseconds) on the amount of CPU time that
205 a process scheduled under a real-time scheduling policy may con‐
206 sume without making a blocking system call. For the purpose of
207 this limit, each time a process makes a blocking system call,
208 the count of its consumed CPU time is reset to zero. The CPU
209 time count is not reset if the process continues trying to use
210 the CPU but is preempted, its time slice expires, or it calls
211 sched_yield(2).
212
213 Upon reaching the soft limit, the process is sent a SIGXCPU sig‐
214 nal. If the process catches or ignores this signal and contin‐
215 ues consuming CPU time, then SIGXCPU will be generated once each
216 second until the hard limit is reached, at which point the
217 process is sent a SIGKILL signal.
218
219 The intended use of this limit is to stop a runaway real-time
220 process from locking up the system.
221
222 For further details on real-time scheduling policies, see
223 sched(7)
224
225 RLIMIT_SIGPENDING (since Linux 2.6.8)
226 This is a limit on the number of signals that may be queued for
227 the real user ID of the calling process. Both standard and
228 real-time signals are counted for the purpose of checking this
229 limit. However, the limit is enforced only for sigqueue(3); it
230 is always possible to use kill(2) to queue one instance of any
231 of the signals that are not already queued to the process.
232
233 RLIMIT_STACK
234 This is the maximum size of the process stack, in bytes. Upon
235 reaching this limit, a SIGSEGV signal is generated. To handle
236 this signal, a process must employ an alternate signal stack
237 (sigaltstack(2)).
238
239 Since Linux 2.6.23, this limit also determines the amount of
240 space used for the process's command-line arguments and environ‐
241 ment variables; for details, see execve(2).
242
243 prlimit()
244 The Linux-specific prlimit() system call combines and extends the func‐
245 tionality of setrlimit() and getrlimit(). It can be used to both set
246 and get the resource limits of an arbitrary process.
247
248 The resource argument has the same meaning as for setrlimit() and getr‐
249 limit().
250
251 If the new_limit argument is not NULL, then the rlimit structure to
252 which it points is used to set new values for the soft and hard limits
253 for resource. If the old_limit argument is not NULL, then a successful
254 call to prlimit() places the previous soft and hard limits for resource
255 in the rlimit structure pointed to by old_limit.
256
257 The pid argument specifies the ID of the process on which the call is
258 to operate. If pid is 0, then the call applies to the calling process.
259 To set or get the resources of a process other than itself, the caller
260 must have the CAP_SYS_RESOURCE capability in the user namespace of the
261 process whose resource limits are being changed, or the real, effec‐
262 tive, and saved set user IDs of the target process must match the real
263 user ID of the caller and the real, effective, and saved set group IDs
264 of the target process must match the real group ID of the caller.
265
267 On success, these system calls return 0. On error, -1 is returned, and
268 errno is set to indicate the error.
269
271 EFAULT A pointer argument points to a location outside the accessible
272 address space.
273
274 EINVAL The value specified in resource is not valid; or, for setr‐
275 limit() or prlimit(): rlim->rlim_cur was greater than
276 rlim->rlim_max.
277
278 EPERM An unprivileged process tried to raise the hard limit; the
279 CAP_SYS_RESOURCE capability is required to do this.
280
281 EPERM The caller tried to increase the hard RLIMIT_NOFILE limit above
282 the maximum defined by /proc/sys/fs/nr_open (see proc(5))
283
284 EPERM (prlimit()) The calling process did not have permission to set
285 limits for the process specified by pid.
286
287 ESRCH Could not find a process with the ID specified in pid.
288
290 For an explanation of the terms used in this section, see at‐
291 tributes(7).
292
293 ┌────────────────────────────────────────────┬───────────────┬─────────┐
294 │Interface │ Attribute │ Value │
295 ├────────────────────────────────────────────┼───────────────┼─────────┤
296 │getrlimit(), setrlimit(), prlimit() │ Thread safety │ MT-Safe │
297 └────────────────────────────────────────────┴───────────────┴─────────┘
298
300 getrlimit()
301 setrlimit()
302 POSIX.1-2008.
303
304 prlimit()
305 Linux.
306
307 RLIMIT_MEMLOCK and RLIMIT_NPROC derive from BSD and are not specified
308 in POSIX.1; they are present on the BSDs and Linux, but on few other
309 implementations. RLIMIT_RSS derives from BSD and is not specified in
310 POSIX.1; it is nevertheless present on most implementations. RLIM‐
311 IT_MSGQUEUE, RLIMIT_NICE, RLIMIT_RTPRIO, RLIMIT_RTTIME, and RLIMIT_SIG‐
312 PENDING are Linux-specific.
313
315 getrlimit()
316 setrlimit()
317 POSIX.1-2001, SVr4, 4.3BSD.
318
319 prlimit()
320 Linux 2.6.36, glibc 2.13.
321
323 A child process created via fork(2) inherits its parent's resource lim‐
324 its. Resource limits are preserved across execve(2).
325
326 Resource limits are per-process attributes that are shared by all of
327 the threads in a process.
328
329 Lowering the soft limit for a resource below the process's current con‐
330 sumption of that resource will succeed (but will prevent the process
331 from further increasing its consumption of the resource).
332
333 One can set the resource limits of the shell using the built-in ulimit
334 command (limit in csh(1)). The shell's resource limits are inherited
335 by the processes that it creates to execute commands.
336
337 Since Linux 2.6.24, the resource limits of any process can be inspected
338 via /proc/pid/limits; see proc(5).
339
340 Ancient systems provided a vlimit() function with a similar purpose to
341 setrlimit(). For backward compatibility, glibc also provides vlimit().
342 All new applications should be written using setrlimit().
343
344 C library/kernel ABI differences
345 Since glibc 2.13, the glibc getrlimit() and setrlimit() wrapper func‐
346 tions no longer invoke the corresponding system calls, but instead em‐
347 ploy prlimit(), for the reasons described in BUGS.
348
349 The name of the glibc wrapper function is prlimit(); the underlying
350 system call is prlimit64().
351
353 In older Linux kernels, the SIGXCPU and SIGKILL signals delivered when
354 a process encountered the soft and hard RLIMIT_CPU limits were deliv‐
355 ered one (CPU) second later than they should have been. This was fixed
356 in Linux 2.6.8.
357
358 In Linux 2.6.x kernels before Linux 2.6.17, a RLIMIT_CPU limit of 0 is
359 wrongly treated as "no limit" (like RLIM_INFINITY). Since Linux
360 2.6.17, setting a limit of 0 does have an effect, but is actually
361 treated as a limit of 1 second.
362
363 A kernel bug means that RLIMIT_RTPRIO does not work in Linux 2.6.12;
364 the problem is fixed in Linux 2.6.13.
365
366 In Linux 2.6.12, there was an off-by-one mismatch between the priority
367 ranges returned by getpriority(2) and RLIMIT_NICE. This had the effect
368 that the actual ceiling for the nice value was calculated as
369 19 - rlim_cur. This was fixed in Linux 2.6.13.
370
371 Since Linux 2.6.12, if a process reaches its soft RLIMIT_CPU limit and
372 has a handler installed for SIGXCPU, then, in addition to invoking the
373 signal handler, the kernel increases the soft limit by one second.
374 This behavior repeats if the process continues to consume CPU time, un‐
375 til the hard limit is reached, at which point the process is killed.
376 Other implementations do not change the RLIMIT_CPU soft limit in this
377 manner, and the Linux behavior is probably not standards conformant;
378 portable applications should avoid relying on this Linux-specific be‐
379 havior. The Linux-specific RLIMIT_RTTIME limit exhibits the same be‐
380 havior when the soft limit is encountered.
381
382 Kernels before Linux 2.4.22 did not diagnose the error EINVAL for setr‐
383 limit() when rlim->rlim_cur was greater than rlim->rlim_max.
384
385 Linux doesn't return an error when an attempt to set RLIMIT_CPU has
386 failed, for compatibility reasons.
387
388 Representation of "large" resource limit values on 32-bit platforms
389 The glibc getrlimit() and setrlimit() wrapper functions use a 64-bit
390 rlim_t data type, even on 32-bit platforms. However, the rlim_t data
391 type used in the getrlimit() and setrlimit() system calls is a (32-bit)
392 unsigned long. Furthermore, in Linux, the kernel represents resource
393 limits on 32-bit platforms as unsigned long. However, a 32-bit data
394 type is not wide enough. The most pertinent limit here is RLIM‐
395 IT_FSIZE, which specifies the maximum size to which a file can grow: to
396 be useful, this limit must be represented using a type that is as wide
397 as the type used to represent file offsets—that is, as wide as a 64-bit
398 off_t (assuming a program compiled with _FILE_OFFSET_BITS=64).
399
400 To work around this kernel limitation, if a program tried to set a re‐
401 source limit to a value larger than can be represented in a 32-bit un‐
402 signed long, then the glibc setrlimit() wrapper function silently con‐
403 verted the limit value to RLIM_INFINITY. In other words, the requested
404 resource limit setting was silently ignored.
405
406 Since glibc 2.13, glibc works around the limitations of the getrlimit()
407 and setrlimit() system calls by implementing setrlimit() and getrlim‐
408 it() as wrapper functions that call prlimit().
409
411 The program below demonstrates the use of prlimit().
412
413 #define _GNU_SOURCE
414 #define _FILE_OFFSET_BITS 64
415 #include <err.h>
416 #include <stdint.h>
417 #include <stdio.h>
418 #include <stdlib.h>
419 #include <sys/resource.h>
420 #include <time.h>
421
422 int
423 main(int argc, char *argv[])
424 {
425 pid_t pid;
426 struct rlimit old, new;
427 struct rlimit *newp;
428
429 if (!(argc == 2 || argc == 4)) {
430 fprintf(stderr, "Usage: %s <pid> [<new-soft-limit> "
431 "<new-hard-limit>]\n", argv[0]);
432 exit(EXIT_FAILURE);
433 }
434
435 pid = atoi(argv[1]); /* PID of target process */
436
437 newp = NULL;
438 if (argc == 4) {
439 new.rlim_cur = atoi(argv[2]);
440 new.rlim_max = atoi(argv[3]);
441 newp = &new;
442 }
443
444 /* Set CPU time limit of target process; retrieve and display
445 previous limit */
446
447 if (prlimit(pid, RLIMIT_CPU, newp, &old) == -1)
448 err(EXIT_FAILURE, "prlimit-1");
449 printf("Previous limits: soft=%jd; hard=%jd\n",
450 (intmax_t) old.rlim_cur, (intmax_t) old.rlim_max);
451
452 /* Retrieve and display new CPU time limit */
453
454 if (prlimit(pid, RLIMIT_CPU, NULL, &old) == -1)
455 err(EXIT_FAILURE, "prlimit-2");
456 printf("New limits: soft=%jd; hard=%jd\n",
457 (intmax_t) old.rlim_cur, (intmax_t) old.rlim_max);
458
459 exit(EXIT_SUCCESS);
460 }
461
463 prlimit(1), dup(2), fcntl(2), fork(2), getrusage(2), mlock(2), mmap(2),
464 open(2), quotactl(2), sbrk(2), shmctl(2), malloc(3), sigqueue(3),
465 ulimit(3), core(5), capabilities(7), cgroups(7), credentials(7), sig‐
466 nal(7)
467
468
469
470Linux man-pages 6.04 2023-03-30 getrlimit(2)