msm_gpu_convert_priority(9)

💅 msm_gpu_convert_priority(9) — from the Kernel API Manual, bestie

okay so real talk, this is one of those tiny little functions that's doing SO much emotional labor behind the scenes. it takes a userspace priority and turns it into a ring number + a scheduler priority. that's it. that's the whole vibe. and it's lowkey iconic for it. 🎀

---

✨ SYNOPSIS ✨

int msm_gpu_convert_priority (struct msm_gpu *gpu , int prio , unsigned
*ring_nr , enum drm_sched_priority *sched_prio );

like yes, that's the whole signature, she's not long, she's efficient, we stan 💅

---

🌸 ARGUMENTS 🌸

• gpu — the gpu instance. your main character. the whole reason we're here 🫶
• prio — the userspace priority level. this is the number the user hands you, and you're about to translate it into something the kernel actually understands
• ring_nr — [out] — the ringbuffer that the userspace priority maps to. it's an output, so you're filling this bad boy in, not reading it 👑
• sched_prio — [out] — the gpu scheduler priority level that the userspace priority maps to. also an output. also yours to populate. no cap ✨

---

🦋 DESCRIPTION 🦋

okay so here's the lore. drm/scheduler provides its own level of prioritization, right? so our total number of available priority levels is literally (nr_rings * NR_SCHED_PRIORITIES). multiply it out, that's the whole menu 🌈

each ring is associated with its own scheduler instance — that's the structural tea ☕

BUT. and this is the plot twist. our UABI says that lower numerical values are higher priority. so mapping that single userspace priority level into ring_nr and sched_prio takes some care, because the directions are kind of opposite-ish and you can't just vibes your way through it 😭

so the userspace provided priority (like, when a submitqueue is created) gets mapped to ring nr and scheduler priority as such:

ring_nr    = userspace_prio / NR_SCHED_PRIORITIES
sched_prio = NR_SCHED_PRIORITIES - (userspace_prio % NR_SCHED_PRIORITIES) - 1

which, honestly? slay. 💅

and why does this matter: it allows generations without preemption (nr_rings==1) to still have some amount of prioritization, and it provides more priority levels for gens that do have preemption. so everybody eats, just at different amounts 🍓

---

🔥 RETURN 🔥

• 0 on success 💖
• -errno on error 💀

---

👑 SEE ALSO 👑

• Kernel file ./drivers/gpu/drm/msm/msm_gpu.h
• struct msm_gpu_devfreq(9)
• struct msm_perfcntr_group_state(9)
• struct msm_perfcntr_state(9)
• struct msm_context(9)
• msm_context_is_vmbind(9)
• struct msm_gpu_submitqueue(9)

---

ngl bestie, if you're touching submitqueue creation, go read struct msm_gpu_submitqueue(9) too. iykyk 🧚

---

and that's the page!! brought to you by ./drivers/gpu/drm/msm, dated October 2026 — mwah, go convert some priorities responsibly, ily, bye!! 💜🦋⭐