amdgpu_hmm_invalidate_hsa(9)

📦 version 7.3.1, in nixpkgs unstable

💅 amdgpu_hmm_invalidate_hsa(9) 💅

okay so bestie, welcome to the Kernel API Manual — this one's a (9) page from the kernel docs, and it's lowkey one of those quiet-but-important callback girlies. no cap.

✨ NAME ✨

amdgpu_hmm_invalidate_hsa — callback to notify about mm change 🫶

🌸 SYNOPSIS 🌸

bool amdgpu_hmm_invalidate_hsa (struct mmu_interval_notifier *mni ,
       const struct mmu_notifier_range *range , unsigned long cur_seq );

it returns a bool, she takes three args, and she's literally the whole vibe. main character energy fr. 👑

🎀 ARGUMENTS 🎀

  • mni — the range (mm) is about to update. yup, this is the struct mmu_interval_notifier * and she is about to get some news she didn't ask for 😭
  • range — details on the invalidation. const struct mmu_notifier_range * — all the deets, all the gossip, wrapped up cute 💌
  • cur_seq — Value to pass to mmu_interval_set_seq. it's unsigned long, and it's exactly the number you hand off to mmu_interval_set_seq. no substitutions, bestie — pass this value. 🍓

🦋 DESCRIPTION 🦋

okay so here's the tea ☕ — we temporarily evict the BO attached to this range. like, moving it out for a sec, not forever, don't panic.

and the way that this necessitates evicting all user-mode queues of the process?? yeah. the entire process's user-mode queues gotta go too. it's giving "one thing leads to another" and honestly the way it's just how it is 😤✨

🔥 SEE ALSO 🔥

like, read the whole friend group, iykyk 💜

ngl this one's short but she matters. go read MMU Notifier(9) if you wanna understand the why, bestie 🧚

and that's the page!! sourced straight from ./drivers/gpu/drm/amd/amdgpu, dated October 2026 — mwah, slay, periodt 💖🌈⭐

📦 where this page comes from