restore_exclusive_pte(9)

💅 restore_exclusive_pte(9) — the Kernel API Manual, bestie edition ✨

okay so you're in the restore_exclusive_pte(9) girlies now, congrats 🎀 this one lives in the Kernel API Manual and it's lowkey a glow-up function: it takes a device-exclusive entry and puts it back to being a normal present pte. iconic behavior honestly 🦋

---

🌸 NAME 🌸

restore_exclusive_pte — restore a device-exclusive entry 💖

---

✨ SYNOPSIS ✨

void restore_exclusive_pte (struct vm_area_struct *vma , struct folio
*folio , struct page *page , unsigned long address , pte_t *ptep , pte_t
orig_pte );

yeah she takes six arguments, she's high-maintenance, we love her for it 💅

---

🫶 ARGUMENTS 🫶

• vma — the VMA that covers address ✨
• folio — the folio that's mapped
• page — the mapped folio page (yes bestie, the page of the folio, not a different thing, they're a set 🎀)
• address — the virtual address, the actual spot, the address-address 🗺️
• ptep — a pte pointer into the locked page table mapping the folio page 🔒
• orig_pte — the pte value sitting at ptep 👑

---

💖 DESCRIPTION 💖

So what she does: restore a device-exclusive non-swap entry back to an ordinary present pte. that's the whole vibe. a little "never mind, put it back" moment 🙈

The rules though, fr fr: the folio and the page table must be locked, and MMU notifiers must have been called to invalidate any (exclusive) device mappings. no shortcuts, no vibes-only locking, it's giving "read the instructions" 📋

Why the folio lock? okay so bestie, locking the folio makes sure that anybody who just converted the pte to a device-exclusive entry can map it into the device to make forward progress without others converting it back until the folio was unlocked. it's a whole "let her finish" situation 👑✨

If the folio lock ever becomes an issue, we can stop relying on the folio lock — it might make some scenarios with heavy thrashing less likely to make forward progress, but these scenarios might not be valid use cases. so like, possible, but also maybe not a real problem, iykyk 🤷‍♀️

⚠️ important caveat time: note that the folio lock does not protect against all cases of concurrent page table modifications (e.g., MADV_DONTNEED, mprotect), so device drivers must use MMU notifiers to sync against any concurrent changes. the folio lock is not the main character here, the MMU notifiers are 💀

---

🔥 SEE ALSO 🔥

Kernel file ./mm/memory.c:

• free_pgd_range(9)
• free_pgtables(9)
• __vm_normal_page(9)
• vm_normal_page(9)
• vm_normal_folio(9)
• vm_normal_page_pmd(9)
• vm_normal_folio_pmd(9)
• vm_normal_page_pud(9)
• cond_install_uffd_wp_ptes(9)
• zap_vma_for_reaping(9)
• unmap_vmas(9)
• zap_vma_range_batched(9)
• zap_vma_range(9)
• zap_special_vma_range(9)
• vm_insert_pages(9)
• vm_insert_page(9)
• vm_map_pages(9)
• vm_map_pages_zero(9)
• vmf_insert_pfn_prot(9)
• vmf_insert_pfn(9)
• remap_pfn_range(9)
• vm_iomap_memory(9)
• __vmf_anon_prepare(9)
• finish_mkwrite_fault(9)
• unmap_mapping_folio(9)
• unmap_mapping_pages(9)
• unmap_mapping_range(9)
• set_pte_range(9)
• finish_fault(9)
• mm_account_fault(9)
• follow_pfnmap_start(9)
• follow_pfnmap_end(9)
• generic_access_phys(9)
• access_remote_vm(9)
• copy_remote_vm_str(9)
• folio_zero_user(9)

like, that's a whole reading list, go off sis 📚💜

---

okay that's the page!! sign-off time 🫶 this beauty came from ./mm, dated October 2026, and she answers to restore_exclusive_pte(9) if you need her again. mwah 💋✨🎀