getrlimit(2)

Purpose

getrlimit() and setrlimit() get and set per-process resource limits; prlimit64() is the modern combined call that can read another process’s limits, set them, or both in one invocation (getrlimit.2, DESCRIPTION). Each resource has a soft limit (what the kernel enforces) and a hard limit (ceiling for the soft limit). This page groups the whole rlimit call family; the resources themselves are listed under “Resources”.

Synopsis

#include <sys/resource.h>
int getrlimit(int resource, struct rlimit *rlim);
int setrlimit(int resource, const struct rlimit *rlim);
int prlimit(pid_t pid, int resource,
            const struct rlimit *new_limit,
            struct rlimit *old_limit);

struct rlimit {
    rlim_t rlim_cur;  /* Soft limit */
    rlim_t rlim_max;  /* Hard limit (ceiling for rlim_cur) */
};

Arguments and behavior

resource selects one of the RLIMIT_* constants (see below). An unprivileged process may raise its soft limit up to the hard limit and may irreversibly lower its hard limit; raising the hard limit or setting another process’s limits requires CAP_SYS_RESOURCE in the initial user namespace (getrlimit.2, DESCRIPTION). RLIM_INFINITY (all-ones rlim_t) means “no limit” in both directions. prlimit() with pid 0 acts on the caller; for a different target the caller needs CAP_SYS_RESOURCE or the same real/effective/saved-set UID and GID as the target, plus ptrace-read access rules described in the man page (getrlimit.2, DESCRIPTION). Either of new_limit/old_limit may be NULL.

Return value and errors

0 on success, -1 with errno (getrlimit.2, RETURN VALUE): EFAULT, EINVAL (bad resource), EPERM (unprivileged caller tried to raise a hard limit or touch a process it may not), and EPERM rules for setrlimit-specific cases (getrlimit.2, ERRORS).

Resources and constants

Values below are the x86_64/glibc enumeration order from /usr/include/x86_64-linux-gnu/bits/resource.h (verified on this host; the kernel’s uapi asm-generic/resource.h uses the same order):

Security and notes

Setuid/setgid transitions must not preserve arbitrary hard limits: the kernel resets several resources for privilege changes (getrlimit.2, NOTES on RLIMIT_* and execve interaction). The history is arch-shaped: asm-generic architectures (arm64, riscv, loongarch) never had getrlimit/setrlimit entry points at all – they call prlimit64 directly, and the ledger’s none numbers are the point, not an omission. On i386/arm32, Linux 2.4 widened the limit type and renumbered the call: the old 16-bit-era slot kept the name getrlimit and the new one landed in slot ugetrlimit (syscalls.2, “Old and new versions” list); glibc’s getrlimit() calls whichever the running kernel prefers and falls back transparently. musl exposes prlimit in sys/resource.h and maps prlimit64 to it with #define prlimit64 prlimit (musl-clone include/sys/resource.h line 58); musl implements getrusage/getrlimit/setrlimit/prlimit/personality in src/misc and src/linux (verified file list in the worklog). See also: getrusage for usage statistics; fork.md for RLIMIT_NPROC enforcement.

References

• Cached man page: source-cache/man-pages-sparse/man2/getrlimit.2; ugetrlimit history in man2/syscalls.2
• Online: https://man7.org/linux/man-pages/man2/getrlimit.2.html
• Header: /usr/include/x86_64-linux-gnu/bits/resource.h (enumeration order), /usr/include/sys/resource.h (prototypes)
• Kernel tables: source-cache/kernel/syscall_64.tbl (97, 160, 302), syscall_32.tbl (76, 75, 340, 191), syscall_generic.tbl (261), syscall_arm.tbl (369, 191)
• musl: source-cache/musl-clone/musl/src/misc/{getrlimit,setrlimit}.c, src/linux/prlimit.c, include/sys/resource.h
• Ledger rows SC-getrlimit, SC-setrlimit, SC-prlimit64, SC-ugetrlimit