1GETRLIMIT(2) Linux Programmer's Manual GETRLIMIT(2)
2
6 getrlimit, setrlimit, prlimit - get/set resource limits
7
9 #include <sys/time.h>
10 #include <sys/resource.h>
11 int getrlimit(int resource, struct rlimit *rlim);
13 int setrlimit(int resource, const struct rlimit *rlim);
14 int prlimit(pid_t pid, int resource, const struct rlimit *new_limit,
16 struct rlimit *old_limit);
17 Feature Test Macro Requirements for glibc (see feature_test_macros(7)):
19 prlimit(): _GNU_SOURCE
21
23 The getrlimit() and setrlimit() system calls get and set resource lim‐
24 its respectively. Each resource has an associated soft and hard limit,
25 as defined by the rlimit structure:
26 struct rlimit {
28 rlim_t rlim_cur; /* Soft limit */
29 rlim_t rlim_max; /* Hard limit (ceiling for rlim_cur) */
30 };
31 The soft limit is the value that the kernel enforces for the corre‐
33 sponding resource. The hard limit acts as a ceiling for the soft
34 limit: an unprivileged process may set only its soft limit to a value
35 in the range from 0 up to the hard limit, and (irreversibly) lower its
36 hard limit. A privileged process (under Linux: one with the
37 CAP_SYS_RESOURCE capability) may make arbitrary changes to either limit
38 value.
39 The value RLIM_INFINITY denotes no limit on a resource (both in the
41 structure returned by getrlimit() and in the structure passed to setr‐
42 limit()).
43 The resource argument must be one of:
45 RLIMIT_AS
47 This is the maximum size of the process's virtual memory
48 (address space). The limit is specified in bytes, and is
49 rounded down to the system page size. This limit affects calls
50 to brk(2), mmap(2), and mremap(2), which fail with the error
51 ENOMEM upon exceeding this limit. In addition, automatic stack
52 expansion fails (and generates a SIGSEGV that kills the process
53 if no alternate stack has been made available via sigalt‐
54 stack(2)). Since the value is a long, on machines with a 32-bit
55 long either this limit is at most 2 GiB, or this resource is
56 unlimited.
57 RLIMIT_CORE
59 This is the maximum size of a core file (see core(5)) in bytes
60 that the process may dump. When 0 no core dump files are cre‐
61 ated. When nonzero, larger dumps are truncated to this size.
62 RLIMIT_CPU
64 This is a limit, in seconds, on the amount of CPU time that the
65 process can consume. When the process reaches the soft limit,
66 it is sent a SIGXCPU signal. The default action for this signal
67 is to terminate the process. However, the signal can be caught,
68 and the handler can return control to the main program. If the
69 process continues to consume CPU time, it will be sent SIGXCPU
70 once per second until the hard limit is reached, at which time
71 it is sent SIGKILL. (This latter point describes Linux behav‐
72 ior. Implementations vary in how they treat processes which
73 continue to consume CPU time after reaching the soft limit.
74 Portable applications that need to catch this signal should per‐
75 form an orderly termination upon first receipt of SIGXCPU.)
76 RLIMIT_DATA
78 This is the maximum size of the process's data segment (initial‐
79 ized data, uninitialized data, and heap). The limit is speci‐
80 fied in bytes, and is rounded down to the system page size.
81 This limit affects calls to brk(2), sbrk(2), and (since Linux
82 4.7) mmap(2), which fail with the error ENOMEM upon encountering
83 the soft limit of this resource.
84 RLIMIT_FSIZE
86 This is the maximum size in bytes of files that the process may
87 create. Attempts to extend a file beyond this limit result in
88 delivery of a SIGXFSZ signal. By default, this signal termi‐
89 nates a process, but a process can catch this signal instead, in
90 which case the relevant system call (e.g., write(2), trun‐
91 cate(2)) fails with the error EFBIG.
92 RLIMIT_LOCKS (early Linux 2.4 only)
94 This is a limit on the combined number of flock(2) locks and
95 fcntl(2) leases that this process may establish.
96 RLIMIT_MEMLOCK
98 This is the maximum number of bytes of memory that may be locked
99 into RAM. This limit is in effect rounded down to the nearest
100 multiple of the system page size. This limit affects mlock(2),
101 mlockall(2), and the mmap(2) MAP_LOCKED operation. Since Linux
102 2.6.9, it also affects the shmctl(2) SHM_LOCK operation, where
103 it sets a maximum on the total bytes in shared memory segments
104 (see shmget(2)) that may be locked by the real user ID of the
105 calling process. The shmctl(2) SHM_LOCK locks are accounted for
106 separately from the per-process memory locks established by
107 mlock(2), mlockall(2), and mmap(2) MAP_LOCKED; a process can
108 lock bytes up to this limit in each of these two categories.
109 In Linux kernels before 2.6.9, this limit controlled the amount
111 of memory that could be locked by a privileged process. Since
112 Linux 2.6.9, no limits are placed on the amount of memory that a
113 privileged process may lock, and this limit instead governs the
114 amount of memory that an unprivileged process may lock.
115 RLIMIT_MSGQUEUE (since Linux 2.6.8)
117 This is a limit on the number of bytes that can be allocated for
118 POSIX message queues for the real user ID of the calling
119 process. This limit is enforced for mq_open(3). Each message
120 queue that the user creates counts (until it is removed) against
121 this limit according to the formula:
122 Since Linux 3.5:
124 bytes = attr.mq_maxmsg * sizeof(struct msg_msg) +
126 min(attr.mq_maxmsg, MQ_PRIO_MAX) *
127 sizeof(struct posix_msg_tree_node)+
128 /* For overhead */
129 attr.mq_maxmsg * attr.mq_msgsize;
130 /* For message data */
131 Linux 3.4 and earlier:
133 bytes = attr.mq_maxmsg * sizeof(struct msg_msg *) +
135 /* For overhead */
136 attr.mq_maxmsg * attr.mq_msgsize;
137 /* For message data */
138 where attr is the mq_attr structure specified as the fourth
140 argument to mq_open(3), and the msg_msg and posix_msg_tree_node
141 structures are kernel-internal structures.
142 The "overhead" addend in the formula accounts for overhead bytes
144 required by the implementation and ensures that the user cannot
145 create an unlimited number of zero-length messages (such mes‐
146 sages nevertheless each consume some system memory for bookkeep‐
147 ing overhead).
148 RLIMIT_NICE (since Linux 2.6.12, but see BUGS below)
150 This specifies a ceiling to which the process's nice value can
151 be raised using setpriority(2) or nice(2). The actual ceiling
152 for the nice value is calculated as 20 - rlim_cur. The useful
153 range for this limit is thus from 1 (corresponding to a nice
154 value of 19) to 40 (corresponding to a nice value of -20). This
155 unusual choice of range was necessary because negative numbers
156 cannot be specified as resource limit values, since they typi‐
157 cally have special meanings. For example, RLIM_INFINITY typi‐
158 cally is the same as -1. For more detail on the nice value, see
159 sched(7).
160 RLIMIT_NOFILE
162 This specifies a value one greater than the maximum file
163 descriptor number that can be opened by this process. Attempts
164 (open(2), pipe(2), dup(2), etc.) to exceed this limit yield the
165 error EMFILE. (Historically, this limit was named RLIMIT_OFILE
166 on BSD.)
167 Since Linux 4.5, this limit also defines the maximum number of
169 file descriptors that an unprivileged process (one without the
170 CAP_SYS_RESOURCE capability) may have "in flight" to other pro‐
171 cesses, by being passed across UNIX domain sockets. This limit
172 applies to the sendmsg(2) system call. For further details, see
173 unix(7).
174 RLIMIT_NPROC
176 This is a limit on the number of extant process (or, more pre‐
177 cisely on Linux, threads) for the real user ID of the calling
178 process. So long as the current number of processes belonging
179 to this process's real user ID is greater than or equal to this
180 limit, fork(2) fails with the error EAGAIN.
181 The RLIMIT_NPROC limit is not enforced for processes that have
183 either the CAP_SYS_ADMIN or the CAP_SYS_RESOURCE capability.
184 RLIMIT_RSS
186 This is a limit (in bytes) on the process's resident set (the
187 number of virtual pages resident in RAM). This limit has effect
188 only in Linux 2.4.x, x < 30, and there affects only calls to
189 madvise(2) specifying MADV_WILLNEED.
190 RLIMIT_RTPRIO (since Linux 2.6.12, but see BUGS)
192 This specifies a ceiling on the real-time priority that may be
193 set for this process using sched_setscheduler(2) and sched_set‐
194 param(2).
195 For further details on real-time scheduling policies, see
197 sched(7)
198 RLIMIT_RTTIME (since Linux 2.6.25)
200 This is a limit (in microseconds) on the amount of CPU time that
201 a process scheduled under a real-time scheduling policy may con‐
202 sume without making a blocking system call. For the purpose of
203 this limit, each time a process makes a blocking system call,
204 the count of its consumed CPU time is reset to zero. The CPU
205 time count is not reset if the process continues trying to use
206 the CPU but is preempted, its time slice expires, or it calls
207 sched_yield(2).
208 Upon reaching the soft limit, the process is sent a SIGXCPU sig‐
210 nal. If the process catches or ignores this signal and contin‐
211 ues consuming CPU time, then SIGXCPU will be generated once each
212 second until the hard limit is reached, at which point the
213 process is sent a SIGKILL signal.
214 The intended use of this limit is to stop a runaway real-time
216 process from locking up the system.
217 For further details on real-time scheduling policies, see
219 sched(7)
220 RLIMIT_SIGPENDING (since Linux 2.6.8)
222 This is a limit on the number of signals that may be queued for
223 the real user ID of the calling process. Both standard and
224 real-time signals are counted for the purpose of checking this
225 limit. However, the limit is enforced only for sigqueue(3); it
226 is always possible to use kill(2) to queue one instance of any
227 of the signals that are not already queued to the process.
228 RLIMIT_STACK
230 This is the maximum size of the process stack, in bytes. Upon
231 reaching this limit, a SIGSEGV signal is generated. To handle
232 this signal, a process must employ an alternate signal stack
233 (sigaltstack(2)).
234 Since Linux 2.6.23, this limit also determines the amount of
236 space used for the process's command-line arguments and environ‐
237 ment variables; for details, see execve(2).
238 prlimit()
240 The Linux-specific prlimit() system call combines and extends the func‐
241 tionality of setrlimit() and getrlimit(). It can be used to both set
242 and get the resource limits of an arbitrary process.
243 The resource argument has the same meaning as for setrlimit() and getr‐
245 limit().
246 If the new_limit argument is a not NULL, then the rlimit structure to
248 which it points is used to set new values for the soft and hard limits
249 for resource. If the old_limit argument is a not NULL, then a success‐
250 ful call to prlimit() places the previous soft and hard limits for
251 resource in the rlimit structure pointed to by old_limit.
252 The pid argument specifies the ID of the process on which the call is
254 to operate. If pid is 0, then the call applies to the calling process.
255 To set or get the resources of a process other than itself, the caller
256 must have the CAP_SYS_RESOURCE capability in the user namespace of the
257 process whose resource limits are being changed, or the real, effec‐
258 tive, and saved set user IDs of the target process must match the real
259 user ID of the caller and the real, effective, and saved set group IDs
260 of the target process must match the real group ID of the caller.
261
263 On success, these system calls return 0. On error, -1 is returned, and
264 errno is set appropriately.
265
267 EFAULT A pointer argument points to a location outside the accessible
268 address space.
269 EINVAL The value specified in resource is not valid; or, for setr‐
271 limit() or prlimit(): rlim->rlim_cur was greater than
272 rlim->rlim_max.
273 EPERM An unprivileged process tried to raise the hard limit; the
275 CAP_SYS_RESOURCE capability is required to do this.
276 EPERM The caller tried to increase the hard RLIMIT_NOFILE limit above
278 the maximum defined by /proc/sys/fs/nr_open (see proc(5))
279 EPERM (prlimit()) The calling process did not have permission to set
281 limits for the process specified by pid.
282 ESRCH Could not find a process with the ID specified in pid.
284
286 The prlimit() system call is available since Linux 2.6.36. Library
287 support is available since glibc 2.13.
288
290 For an explanation of the terms used in this section, see
291 attributes(7).
292 ┌────────────────────────────────────┬───────────────┬─────────┐
294 │Interface │ Attribute │ Value │
295 ├────────────────────────────────────┼───────────────┼─────────┤
296 │getrlimit(), setrlimit(), prlimit() │ Thread safety │ MT-Safe │
297 └────────────────────────────────────┴───────────────┴─────────┘
298
300 getrlimit(), setrlimit(): POSIX.1-2001, POSIX.1-2008, SVr4, 4.3BSD.
301 prlimit(): Linux-specific.
303 RLIMIT_MEMLOCK and RLIMIT_NPROC derive from BSD and are not specified
305 in POSIX.1; they are present on the BSDs and Linux, but on few other
306 implementations. RLIMIT_RSS derives from BSD and is not specified in
307 POSIX.1; it is nevertheless present on most implementations.
308 RLIMIT_MSGQUEUE, RLIMIT_NICE, RLIMIT_RTPRIO, RLIMIT_RTTIME, and
309 RLIMIT_SIGPENDING are Linux-specific.
310
312 A child process created via fork(2) inherits its parent's resource lim‐
313 its. Resource limits are preserved across execve(2).
314 Lowering the soft limit for a resource below the process's current con‐
316 sumption of that resource will succeed (but will prevent the process
317 from further increasing its consumption of the resource).
318 One can set the resource limits of the shell using the built-in ulimit
320 command (limit in csh(1)). The shell's resource limits are inherited
321 by the processes that it creates to execute commands.
322 Since Linux 2.6.24, the resource limits of any process can be inspected
324 via /proc/[pid]/limits; see proc(5).
325 Ancient systems provided a vlimit() function with a similar purpose to
327 setrlimit(). For backward compatibility, glibc also provides vlimit().
328 All new applications should be written using setrlimit().
329 C library/kernel ABI differences
331 Since version 2.13, the glibc getrlimit() and setrlimit() wrapper func‐
332 tions no longer invoke the corresponding system calls, but instead
333 employ prlimit(), for the reasons described in BUGS.
334 The name of the glibc wrapper function is prlimit(); the underlying
336 system call is prlimit64().
337
339 In older Linux kernels, the SIGXCPU and SIGKILL signals delivered when
340 a process encountered the soft and hard RLIMIT_CPU limits were deliv‐
341 ered one (CPU) second later than they should have been. This was fixed
342 in kernel 2.6.8.
343 In 2.6.x kernels before 2.6.17, a RLIMIT_CPU limit of 0 is wrongly
345 treated as "no limit" (like RLIM_INFINITY). Since Linux 2.6.17, set‐
346 ting a limit of 0 does have an effect, but is actually treated as a
347 limit of 1 second.
348 A kernel bug means that RLIMIT_RTPRIO does not work in kernel 2.6.12;
350 the problem is fixed in kernel 2.6.13.
351 In kernel 2.6.12, there was an off-by-one mismatch between the priority
353 ranges returned by getpriority(2) and RLIMIT_NICE. This had the effect
354 that the actual ceiling for the nice value was calculated as
355 19 - rlim_cur. This was fixed in kernel 2.6.13.
356 Since Linux 2.6.12, if a process reaches its soft RLIMIT_CPU limit and
358 has a handler installed for SIGXCPU, then, in addition to invoking the
359 signal handler, the kernel increases the soft limit by one second.
360 This behavior repeats if the process continues to consume CPU time,
361 until the hard limit is reached, at which point the process is killed.
362 Other implementations do not change the RLIMIT_CPU soft limit in this
363 manner, and the Linux behavior is probably not standards conformant;
364 portable applications should avoid relying on this Linux-specific
365 behavior. The Linux-specific RLIMIT_RTTIME limit exhibits the same
366 behavior when the soft limit is encountered.
367 Kernels before 2.4.22 did not diagnose the error EINVAL for setrlimit()
369 when rlim->rlim_cur was greater than rlim->rlim_max.
370 Representation of "large" resource limit values on 32-bit platforms
372 The glibc getrlimit() and setrlimit() wrapper functions use a 64-bit
373 rlim_t data type, even on 32-bit platforms. However, the rlim_t data
374 type used in the getrlimit() and setrlimit() system calls is a (32-bit)
375 unsigned long. Furthermore, in Linux versions before 2.6.36, the ker‐
376 nel represents resource limits on 32-bit platforms as unsigned long.
377 However, a 32-bit data type is not wide enough. The most pertinent
378 limit here is RLIMIT_FSIZE, which specifies the maximum size to which a
379 file can grow: to be useful, this limit must be represented using a
380 type that is as wide as the type used to represent file offsets—that
381 is, as wide as a 64-bit off_t (assuming a program compiled with
382 _FILE_OFFSET_BITS=64).
383 To work around this kernel limitation, if a program tried to set a
385 resource limit to a value larger than can be represented in a 32-bit
386 unsigned long, then the glibc setrlimit() wrapper function silently
387 converted the limit value to RLIM_INFINITY. In other words, the
388 requested resource limit setting was silently ignored.
389 This problem was addressed in Linux 2.6.36 with two principal changes:
391 * the addition of a new kernel representation of resource limits that
393 uses 64 bits, even on 32-bit platforms;
394 * the addition of the prlimit() system call, which employs 64-bit val‐
396 ues for its resource limit arguments.
397 Since version 2.13, glibc works around the limitations of the getr‐
399 limit() and setrlimit() system calls by implementing setrlimit() and
400 getrlimit() as wrapper functions that call prlimit().
401
403 The program below demonstrates the use of prlimit().
404 #define _GNU_SOURCE
406 #define _FILE_OFFSET_BITS 64
407 #include <stdio.h>
408 #include <time.h>
409 #include <stdlib.h>
410 #include <unistd.h>
411 #include <sys/resource.h>
412 #define errExit(msg) do { perror(msg); exit(EXIT_FAILURE); \
414 } while (0)
415 int
417 main(int argc, char *argv[])
418 {
419 struct rlimit old, new;
420 struct rlimit *newp;
421 pid_t pid;
422 if (!(argc == 2 || argc == 4)) {
424 fprintf(stderr, "Usage: %s <pid> [<new-soft-limit> "
425 "<new-hard-limit>]\n", argv[0]);
426 exit(EXIT_FAILURE);
427 }
428 pid = atoi(argv[1]); /* PID of target process */
430 newp = NULL;
432 if (argc == 4) {
433 new.rlim_cur = atoi(argv[2]);
434 new.rlim_max = atoi(argv[3]);
435 newp = &new;
436 }
437 /* Set CPU time limit of target process; retrieve and display
439 previous limit */
440 if (prlimit(pid, RLIMIT_CPU, newp, &old) == -1)
442 errExit("prlimit-1");
443 printf("Previous limits: soft=%lld; hard=%lld\n",
444 (long long) old.rlim_cur, (long long) old.rlim_max);
445 /* Retrieve and display new CPU time limit */
447 if (prlimit(pid, RLIMIT_CPU, NULL, &old) == -1)
449 errExit("prlimit-2");
450 printf("New limits: soft=%lld; hard=%lld\n",
451 (long long) old.rlim_cur, (long long) old.rlim_max);
452 exit(EXIT_SUCCESS);
454 }
455
457 prlimit(1), dup(2), fcntl(2), fork(2), getrusage(2), mlock(2), mmap(2),
458 open(2), quotactl(2), sbrk(2), shmctl(2), malloc(3), sigqueue(3),
459 ulimit(3), core(5), capabilities(7), cgroups(7), credentials(7), sig‐
460 nal(7)
461
463 This page is part of release 4.15 of the Linux man-pages project. A
464 description of the project, information about reporting bugs, and the
465 latest version of this page, can be found at
466 https://www.kernel.org/doc/man-pages/.
467Linux 2017-09-15 GETRLIMIT(2)