sched_setscheduler(2)

💖 sched_setscheduler(2) — the sched_setscheduler / sched_getscheduler girlies 💖

okay bestie, pull up a chair, because today we're talking about the system calls that decide which thread gets to run and how it gets to run. these live in section 2 of the manual (system calls, iykyk), and they are literally the main character energy of the Linux scheduler. 🎀

---

✨ NAME ✨

<pink>sched_setscheduler</pink>, <pink>sched_getscheduler</pink> — set and get scheduling policy/parameters. slay, that's the whole job. 💅

🌸 LIBRARY 🌸

Standard C library (<purple>libc</purple>, <purple>-lc</purple>)

✨ SYNOPSIS ✨

#include <sched.h>

int sched_setscheduler(pid_t pid, int policy,
                       const struct sched_param *param);
int sched_getscheduler(pid_t pid);

🦋 DESCRIPTION 🦋

so. the <cyan>sched_setscheduler()</cyan> system call is the one that sets both the scheduling policy and the parameters for the thread whose ID you pass in as pid. multitasking queen. 👑

and if pid equals zero? cute. then it sets the scheduling policy and parameters of the calling thread itself. self-care. 🌸

the scheduling parameters go in the param argument, which is a pointer to a structure shaped like this:

struct sched_param {
    ...
    int sched_priority;
    ...
};

ngl, in the current implementation the structure contains only one field, <purple>sched_priority</purple>. and the way param gets interpreted totally depends on the policy you picked. like, it's giving "context matters." 💅

---

🌈 the "normal" (i.e., non-real-time) policies 🌈

currently, Linux supports the following "normal" — i.e. non-real-time — scheduling policies as values you can shove into policy:

• <mint>SCHED_OTHER</mint> — the standard round-robin time-sharing policy. the default bestie, the people's princess. ✨
• <mint>SCHED_BATCH</mint> — for "batch" style execution of processes. lowkey efficient. 🍓
• <mint>SCHED_IDLE</mint> — for running very low priority background jobs. it's giving "no thoughts, head empty, run whenever." 💀

for each of the policies above, param->sched_priority must be 0. no exceptions, bestie. 🚨

---

🔥 the "real-time" policies 🔥

various "real-time" policies are also supported, for special time-critical applications that need precise control over the way runnable threads get selected for execution. this is the VIP section, no cap. 👑

for the rules governing when a process may use these policies, go see <purple>sched(7)</purple>. the real-time policies you can specify in policy are:

• <hotpink>SCHED_FIFO</hotpink> — a first-in, first-out policy; and
• <hotpink>SCHED_RR</hotpink> — a round-robin policy.

for each of the above, param->sched_priority specifies a scheduling priority for the thread. 🎀 this is a number in the range returned by calling <purple>sched_get_priority_min(2)</purple> and <purple>sched_get_priority_max(2)</purple> with the specified policy. and on Linux? those two system calls return, respectively, <yellow>1</yellow> and <yellow>99</yellow>. periodt. 💅

---

✨ the SCHED_RESET_ON_FORK moment ✨

since Linux <yellow>2.6.32</yellow>, the <mint>SCHED_RESET_ON_FORK</mint> flag can be ORed in policy when you call sched_setscheduler(). what does including this flag do? children created by <purple>fork(2)</purple> do not inherit privileged scheduling policies. iconic behaviour tbh. see <purple>sched(7)</purple> for the details. 🫶

---

🐛 and now, sched_getscheduler() 🐛

<cyan>sched_getscheduler()</cyan> returns the current scheduling policy of the thread identified by pid. and yep — if pid equals zero, the policy of the calling thread will be retrieved. it's the same trick, just backwards. 🦋

⭐ RETURN VALUE ⭐

• on success, <cyan>sched_setscheduler()</cyan> returns <yellow>zero</yellow>. 🎉
• on success, <cyan>sched_getscheduler()</cyan> returns the policy for the thread (a nonnegative integer). ✨
• on error, both calls return <red>-1</red>, and <purple>errno</purple> is set to indicate the error. 😭

💀 ERRORS 💀

• EINVAL — Invalid arguments: pid is negative or param is NULL. 🙈
• EINVAL — (sched_setscheduler()) policy is not one of the recognized policies. like, be serious. 💅
• EINVAL — (sched_setscheduler()) param does not make sense for the specified policy. mismatch energy fr fr. 🚫
• EPERM — The calling thread does not have appropriate privileges. girl, you don't have clearance for this VIP section. 👑
• ESRCH — The thread whose ID is pid could not be found. she's not here, bestie. 💀

🌈 VERSIONS 🌈

POSIX.1 does not detail the permissions that an unprivileged thread requires in order to call <cyan>sched_setscheduler()</cyan>, and the details vary across systems. it's giving "every OS has its own rules." for example, the Solaris 7 manual page says that the real or effective user ID of the caller must match the real user ID or the save set-user-ID of the target. ngl, the drama.

the scheduling policy and parameters are in fact per-thread attributes on Linux. 🧚 that means:

• the value returned from a call to <purple>gettid(2)</purple> can be passed in the argument pid.
• specifying pid as <yellow>0</yellow> will operate on the attributes of the calling thread.
• passing the value returned from a call to <purple>getpid(2)</purple> will operate on the attributes of the main thread of the thread group.

(if you are using the POSIX threads API, then use <purple>pthread_setschedparam(3)</purple>, <purple>pthread_getschedparam(3)</purple>, and <purple>pthread_setschedprio(3)</purple>, instead of the sched_*(2) system calls. that's the assigned reading, bestie. 📚)

🌸 STANDARDS 🌸

<mint>POSIX.1-2024</mint> (but see BUGS below 👀).

<mint>SCHED_BATCH</mint> and <mint>SCHED_IDLE</mint> are Linux-specific. 💅

🎀 HISTORY 🎀

<mint>POSIX.1-2001</mint>. classic. vintage. we love an OG.

⭐ NOTES ⭐

further details of the semantics of all of the above "normal" and "real-time" scheduling policies can be found in the <purple>sched(7)</purple> manual page. that page also describes an additional policy, <hotpink>SCHED_DEADLINE</hotpink>, which is settable only via <purple>sched_setattr(2)</purple>. exclusive fr. 🔥

POSIX systems on which <cyan>sched_setscheduler()</cyan> and <cyan>sched_getscheduler()</cyan> are available define <purple>_POSIX_PRIORITY_SCHEDULING</purple> in <purple><unistd.h></purple>. check for it, bestie. 🦋

🚨 BUGS 🚨

okay this is the unhinged part. POSIX.1 says that on success, sched_setscheduler() should return the previous scheduling policy. but Linux sched_setscheduler() does not conform to this requirement — because it always returns <yellow>0</yellow> on success. 💀 she simply refuses. that's the tea.

💖 SEE ALSO 💖

<purple>chrt(1)</purple>, <purple>nice(2)</purple>, <purple>sched_get_priority_max(2)</purple>, <purple>sched_get_priority_min(2)</purple>, <purple>sched_getaffinity(2)</purple>, <purple>sched_getattr(2)</purple>, <purple>sched_getparam(2)</purple>, <purple>sched_rr_get_interval(2)</purple>, <purple>sched_setaffinity(2)</purple>, <purple>sched_setattr(2)</purple>, <purple>sched_setparam(2)</purple>, <purple>sched_yield(2)</purple>, <purple>setpriority(2)</purple>, <purple>capabilities(7)</purple>, <purple>cpuset(7)</purple>, <purple>sched(7)</purple>

---

and that's the whole page, bestie!! 🫶✨ this little love letter is from Linux man-pages 6.20, dated 2026-02-08 — shoutout to the man-pages girlies for keeping section 2 so cunty. now go set some priorities (responsibly!!). 💅🎀💖