update_tg_load_avg(9)

💖 update_tg_load_avg(9) 💖

okay so bestie, we're deep in the Kernel API Manual today — yes, the actual kernel, yes, the scheduler, don't panic, I got you 🫶 this one is a little CFS scheduling function and it's honestly not that deep once you see what it's doing. let's go ✨

---

🌸 NAME 🌸

update_tg_load_avg — update the tg's load avg. that's it, that's the whole name, and honestly it ate 💅

---

🎀 SYNOPSIS 🎀

void update_tg_load_avg (struct cfs_rq *cfs_rq );

it returns void, so like... don't go looking for a return value, there isn't one, she gives nothing back, she just does 💅

---

🦋 ARGUMENTS 🦋

• cfs_rq — the cfs_rq whose avg changed. that's the one you're handing over, the one that moved, the main character of this call 🌟

---

✨ DESCRIPTION ✨

so this function 'ensures' — and yes the manpage itself put 'ensures' in quotes like that, iconic of it — that:

tg->load_avg :=  tg->cfs_rq[]->avg.load

but here's the plot twist 😭 because tg->load_avg is a global value, there are performance considerations. like, real ones. the vibe check says: don't be out here doing expensive things in a hot path, bestie.

so how do we dodge that? in order to avoid having to look at the other cfs_rq's, we use a differential update where we store the last value we propagated 👑 that's the whole trick.

and this in turn allows skipping updates if the differential is 'small' — again with the quotes, and again, that's literally what the page says, I'm not making this up 🙈 so yeah, small diff = we can skip, and that's a slay for your CPU cycles fr fr.

⭐ key thing to remember: updating tg's load_avg is necessary before update_cfs_group. not optional, not "if you feel like it", it's a prerequisite. do it first, periodt.

---

🌈 SEE ALSO 🌈

• Kernel file ./kernel/sched/fair.c — like, the actual source, go read it, iykyk 📖
• update_cfs_rq_load_avg(9)
• attach_entity_load_avg(9)
• detach_entity_load_avg(9)
• cpu_util(9)
• sched_use_asym_prio(9)
• sched_group_asym(9)
• update_sg_lb_stats(9)
• update_sd_pick_busiest(9)
• idle_cpu_without(9)
• update_sd_lb_stats(9)
• calculate_imbalance(9)
• sched_balance_find_src_group(9)

---

and that's the whole page bestie, you're officially caught up on load average propagation 🧚💜 this came to you from the ./kernel/sched directory, and the page is dated October 2026 — so yeah, future-coded of it, we love a timestamp with range 🌸 bye, go read update_cfs_group next, mwah 😭💖