xe_vm_range_unbind(9)

💅 xe_vm_range_unbind(9) — or, "how to say bye to a VM range" 🎀

okay bestie, buckle up, because we're diving into the Kernel API Manual and this one's a little function with a big job. it lives in xe_vm_range_unbind(9) and honestly? it's giving "cleanup crew, but make it GPU" ✨

---

🌸 NAME 🌸

xe_vm_range_unbind — VM range unbind. that's it, that's the whole vibe. short, sweet, and it does exactly what it says on the tin 💖

---

✨ SYNOPSIS ✨

here's the signature, character for character, no cap:

struct dma_fence * xe_vm_range_unbind (struct xe_vm *vm , struct
       xe_svm_range *range );

so you hand it a vm and a range, and it hands you back a struct dma_fence *. iconic behavior 🫶

---

🦋 ARGUMENTS 🦋

two lil' parameters, both essential, neither optional:

• vm — the VM which the range belongs to. it's the range's home base, its main character, the whole reason we're here 💫
• range — the SVM range to rebind. (yes, the page says rebind even though the function is called unbind — we don't make the rules, we just document them 🙈)

---

💜 DESCRIPTION 💜

okay so here's what actually goes down: this bad boy unbinds an SVM range, which removes the GPU page tables for that range. poof 💨 gone. the page tables said "not today" and left the chat.

literally it's the cleanup step. the "you don't live here anymore" moment. and honestly? slay 🔥

---

⭐ RETURN ⭐

what do you get back, you ask? glad you asked bestie:

• on success — a dma fence for the unbind, so you can signal completion. yes queen, asynchronous and polite 💅
• on failure — an ERR_PTR. 😭 that's your signal that something went sideways, check it before you dereference, fr fr.

---

🌈 SEE ALSO 🌈

okay this is the part where the manpage goes FULL group-chat-adds-everyone mode. the kernel file in question is:

./drivers/gpu/drm/xe/xe_vm.c

and here's the whole squad, no one left behind, we're all in this thread 🧚:

• xe_vm_drm_exec_lock(9)
• xe_vm_remove_compute_exec_queue(9)
• xe_vm_kill(9)
• xe_vm_validate_rebind(9)
• xe_vm_bump_lru(9)
• xe_vm_resume_rebind_worker(9)
• xe_vm_add_fault_entry_pf(9)
• xe_vm_range_rebind(9)
• xe_vma_mem_attr_copy(9)
• xe_vm_lock_vma(9)
• xe_vm_create_scratch(9)
• xe_vm_find_vma_by_addr(9)
• xe_vma_has_default_mem_attrs(9)
• struct xe_vma_lock_and_validate_flags(9)
• xe_vm_bind_kernel_bo(9)
• xe_vm_lock(9)
• xe_vm_unlock(9)
• xe_vm_invalidate_vma_submit(9)
• xe_vm_invalidate_vma(9)
• xe_vma_need_vram_for_atomic(9)
• xe_vm_alloc_madvise_vma(9)
• xe_vm_find_cpu_addr_mirror_vma_range(9)
• xe_vm_alloc_cpu_addr_mirror_vma(9)
• xe_vm_add_exec_queue(9)
• xe_vm_remove_exec_queue(9)

like I said. everyone. and yes, struct xe_vma_lock_and_validate_flags(9) is written exactly like that — we're not touching it, we're not questioning it, we're just passing it along 💀

---

pro tip from me to you 🫶 — xe_vm_range_rebind(9) is right there in the list, so if you ever unbind and then get the urge to rebind... the kernel already thought of that. iykyk ✨

---

and that's the page, bestie 🎀 this gem came out of ./drivers/gpu/drm/xe, dated October 2026, and it's filed under xe_vm_range_unbind(9) in the Kernel API Manual. go forth, unbind responsibly, and remember: the page tables were never yours to keep 🌸💖