On Tue, 22 Sep 2026 16:35:31 +0100 "Lorenzo Stoakes (ARM)" <[email protected]>
wrote:
> The majority of architectures in the kernel defer page table freeing until
> an RCU grace period has elapsed, this series converts all remaining
> architectures to do so too and eliminates CONFIG_MMU_GATHER_RCU_TABLE_FREE
> altogether.
>
> This is important because it enables safe lockless page table walking under
> RCU alone.
>
> Doing so allows for reduced lock contention, avoids lock ordering concerns
> and enables fast, efficient and correct page table walking as a result.
>
> Additionally it removes a bunch of code and architecture-specific behaviour
> which is always a beneficial thing to do.
>
> There has been much recent work on this:
Thanks, I've updated mm.git's mm-unstable branch to this version. Plus
one -fix for [01/12].
> v4:
> * Updated the 1st patch to allocate a new page table on PTE deposit rather
> than deposit a page table that might currently be being walked by an
> RCU-only page table walker, as per David.
Here's how v4 altered mm.git:
mm/huge_memory.c | 2 +-
mm/khugepaged.c | 34 ++++++++++++++++++++++++++++++++--
2 files changed, 33 insertions(+), 3 deletions(-)
--- a/mm/huge_memory.c~b
+++ a/mm/huge_memory.c
@@ -2478,7 +2478,7 @@ static inline void zap_deposited_table(s
pgtable_t pgtable;
pgtable = pgtable_trans_huge_withdraw(mm, pmd);
- pte_free_defer(mm, pgtable);
+ pte_free(mm, pgtable);
mm_dec_nr_ptes(mm);
}
--- a/mm/khugepaged.c~b
+++ a/mm/khugepaged.c
@@ -1278,6 +1278,23 @@ static enum scan_result alloc_charge_fol
return SCAN_SUCCEED;
}
+static pgtable_t alloc_deposit_pte(struct mm_struct *mm)
+{
+ /*
+ * khugepaged is run from a kernel thread, so need to manually set the
+ * correct memcg so the allocation gets charged correctly.
+ */
+ struct mem_cgroup *memcg = get_mem_cgroup_from_mm(mm);
+ struct mem_cgroup *old_memcg = set_active_memcg(memcg);
+ pgtable_t pgtable;
+
+ pgtable = pte_alloc_one(mm);
+
+ set_active_memcg(old_memcg);
+ mem_cgroup_put(memcg);
+ return pgtable;
+}
+
/*
* collapse_huge_page() expects the mmap_lock to be unlocked before entering
and
* will always return with the lock unlocked, to avoid holding the mmap_lock
@@ -1293,7 +1310,7 @@ static enum scan_result collapse_huge_pa
LIST_HEAD(compound_pagelist);
pmd_t *pmd, _pmd;
pte_t *pte = NULL;
- pgtable_t pgtable;
+ pgtable_t pgtable = NULL;
struct folio *folio;
spinlock_t *pmd_ptl, *pte_ptl;
enum scan_result result = SCAN_FAIL;
@@ -1310,6 +1327,14 @@ static enum scan_result collapse_huge_pa
goto out_nolock;
}
+ if (is_pmd_order(order)) {
+ pgtable = alloc_deposit_pte(mm);
+ if (!pgtable) {
+ result = SCAN_ALLOC_HUGE_PAGE_FAIL;
+ goto out_nolock;
+ }
+ }
+
mmap_read_lock(mm);
result = hugepage_vma_revalidate(mm, pmd_addr, /*expect_anon=*/ true,
&vma, cc, order);
@@ -1433,8 +1458,8 @@ static enum scan_result collapse_huge_pa
spin_lock(pmd_ptl);
VM_WARN_ON_ONCE(!pmd_none(*pmd));
if (is_pmd_order(order)) {
- pgtable = pmd_pgtable(_pmd);
pgtable_trans_huge_deposit(mm, pmd, pgtable);
+ pgtable = NULL;
map_anon_folio_pmd_nopf(folio, pmd, vma, pmd_addr);
} else {
/*
@@ -1453,6 +1478,9 @@ static enum scan_result collapse_huge_pa
}
spin_unlock(pmd_ptl);
+ if (is_pmd_order(order))
+ pte_free_defer(mm, pmd_pgtable(_pmd));
+
folio = NULL;
result = SCAN_SUCCEED;
@@ -1463,6 +1491,8 @@ out_up_write:
anon_vma_unlock_write(vma->anon_vma);
mmap_write_unlock(mm);
out_nolock:
+ if (pgtable)
+ pte_free(mm, pgtable);
if (folio)
folio_put(folio);
trace_mm_collapse_huge_page(mm, result == SCAN_SUCCEED, result, order);
_
