Re: [RFC v2 00/21] mm: thp: lazy PTE page table allocation at PMD split

[Nico Pache <npache@redhat.com>]

On Thu, Feb 26, 2026 at 4:33 AM Usama Arif <usama.arif@linux.dev> wrote:
>
> When the kernel creates a PMD-level THP mapping for anonymous pages, it
> pre-allocates a PTE page table via pgtable_trans_huge_deposit(). This
> page table sits unused in a deposit list for the lifetime of the THP
> mapping, only to be withdrawn when the PMD is split or zapped. Every
> anonymous THP therefore wastes 4KB of memory unconditionally. On large
> servers where hundreds of gigabytes of memory are mapped as THPs, this
> adds up: roughly 200MB wasted per 100GB of THP memory. This memory
> could otherwise satisfy other allocations, including the very PTE page
> table allocations needed when splits eventually occur.
>
> This series removes the pre-deposit and allocates the PTE page table
> lazily — only when a PMD split actually happens. Since a large number
> of THPs are never split (they are zapped wholesale when processes exit or
> munmap the full range), the allocation is avoided entirely in the common
> case.
>
> The pre-deposit pattern exists because split_huge_pmd was designed as an
> operation that must never fail: if the kernel decides to split, it needs
> a PTE page table, so one is deposited in advance. But "must never fail"
> is an unnecessarily strong requirement. A PMD split is typically triggered
> by a partial operation on a sub-PMD range — partial munmap, partial
> mprotect, partial mremap and so on.
> Most of these operations already have well-defined error handling for
> allocation failures (e.g., -ENOMEM, VM_FAULT_OOM). Allowing split to
> fail and propagating the error through these existing paths is the natural
> thing to do. Furthermore, split failing requires an order-0 allocation for
> a page table to fail, which is extremely unlikely.
>
> Designing functions like split_huge_pmd as operations that cannot fail
> has a subtle but real cost to code quality. It forces a pre-allocation
> pattern - every THP creation path must deposit a page table, and every
> split or zap path must withdraw one, creating a hidden coupling between
> widely separated code paths.
>
> This also serves as a code cleanup. On every architecture except powerpc
> with hash MMU, the deposit/withdraw machinery becomes dead code. The
> series removes the generic implementations in pgtable-generic.c and the
> s390/sparc overrides, replacing them with no-op stubs guarded by
> arch_needs_pgtable_deposit(), which evaluates to false at compile time
> on all non-powerpc architectures.

Hi Usama,

Thanks for tackling this, it seems like an interesting problem. Im
trying to get more into reviewing, so bare with me I may have some
stupid comments or questions. Where I can really help out is with
testing. I will build this for all RH-supported architectures and run
some automated test suites and performance metrics. I'll report back
if I spot anything.

Cheers!
-- Nico

>
> The series is structured as follows:
>
> Patches 1-2:    Error infrastructure — make split functions return int
>                 and propagate errors from vma_adjust_trans_huge()
>                 through __split_vma, vma_shrink, and commit_merge.
>
> Patches 3-12:   Handle split failure at every call site — copy_huge_pmd,
>                 do_huge_pmd_wp_page, zap_pmd_range, wp_huge_pmd,
>                 change_pmd_range (mprotect), follow_pmd_mask (GUP),
>                 walk_pmd_range (pagewalk), move_page_tables (mremap),
>                 move_pages (userfaultfd), and device migration.
>                 The code will become affective in Patch 14 when split
>                 functions start returning -ENOMEM.
>
> Patch 13:       Add __must_check to __split_huge_pmd(), split_huge_pmd()
>                 and split_huge_pmd_address() so the compiler warns on
>                 unchecked return values.
>
> Patch 14:       The actual change — allocate PTE page tables lazily at
>                 split time instead of pre-depositing at THP creation.
>                 This is when split functions will actually start returning
>                 -ENOMEM.
>
> Patch 15:       Remove the now-dead deposit/withdraw code on
>                 non-powerpc architectures.
>
> Patch 16:       Add THP_SPLIT_PMD_FAILED vmstat counter for monitoring
>                 split failures.
>
> Patches 17-21:  Selftests covering partial munmap, mprotect, mlock,
>                 mremap, and MADV_DONTNEED on THPs to exercise the
>                 split paths.
>
> The error handling patches are placed before the lazy allocation patch so
> that every call site is already prepared to handle split failures before
> the failure mode is introduced. This makes each patch independently safe
> to apply and bisect through.
>
> The patches were tested with CONFIG_DEBUG_ATOMIC_SLEEP and CONFIG_DEBUG_VM
> enabled. The test results are below:
>
> TAP version 13
> 1..5
> # Starting 5 tests from 1 test cases.
> #  RUN           thp_pmd_split.partial_munmap ...
> # thp_pmd_split_test.c:60:partial_munmap:thp_split_pmd: 0 -> 1
> # thp_pmd_split_test.c:62:partial_munmap:thp_split_pmd_failed: 0 -> 0
> #            OK  thp_pmd_split.partial_munmap
> ok 1 thp_pmd_split.partial_munmap
> #  RUN           thp_pmd_split.partial_mprotect ...
> # thp_pmd_split_test.c:60:partial_mprotect:thp_split_pmd: 1 -> 2
> # thp_pmd_split_test.c:62:partial_mprotect:thp_split_pmd_failed: 0 -> 0
> #            OK  thp_pmd_split.partial_mprotect
> ok 2 thp_pmd_split.partial_mprotect
> #  RUN           thp_pmd_split.partial_mlock ...
> # thp_pmd_split_test.c:60:partial_mlock:thp_split_pmd: 2 -> 3
> # thp_pmd_split_test.c:62:partial_mlock:thp_split_pmd_failed: 0 -> 0
> #            OK  thp_pmd_split.partial_mlock
> ok 3 thp_pmd_split.partial_mlock
> #  RUN           thp_pmd_split.partial_mremap ...
> # thp_pmd_split_test.c:60:partial_mremap:thp_split_pmd: 3 -> 4
> # thp_pmd_split_test.c:62:partial_mremap:thp_split_pmd_failed: 0 -> 0
> #            OK  thp_pmd_split.partial_mremap
> ok 4 thp_pmd_split.partial_mremap
> #  RUN           thp_pmd_split.partial_madv_dontneed ...
> # thp_pmd_split_test.c:60:partial_madv_dontneed:thp_split_pmd: 4 -> 5
> # thp_pmd_split_test.c:62:partial_madv_dontneed:thp_split_pmd_failed: 0 -> 0
> #            OK  thp_pmd_split.partial_madv_dontneed
> ok 5 thp_pmd_split.partial_madv_dontneed
> # PASSED: 5 / 5 tests passed.
> # Totals: pass:5 fail:0 xfail:0 xpass:0 skip:0 error:0
>
> The patches are based off of 957a3fab8811b455420128ea5f41c51fd23eb6c7 from
> mm-unstable as of 25 Feb (7.0.0-rc1).
>
>
> RFC v1 -> v2: https://lore.kernel.org/all/20260211125507.4175026-1-usama.arif@linux.dev/
> - Change counter name to THP_SPLIT_PMD_FAILED (David)
> - remove pgtable_trans_huge_{deposit/withdraw} when not needed and
>   make them arch specific (David)
> - make split functions return error code and have callers handle them
>   (David and Kiryl)
> - Add test cases for splitting
>
> Usama Arif (21):
>   mm: thp: make split_huge_pmd functions return int for error
>     propagation
>   mm: thp: propagate split failure from vma_adjust_trans_huge()
>   mm: thp: handle split failure in copy_huge_pmd()
>   mm: thp: handle split failure in do_huge_pmd_wp_page()
>   mm: thp: handle split failure in zap_pmd_range()
>   mm: thp: handle split failure in wp_huge_pmd()
>   mm: thp: retry on split failure in change_pmd_range()
>   mm: thp: handle split failure in follow_pmd_mask()
>   mm: handle walk_page_range() failure from THP split
>   mm: thp: handle split failure in mremap move_page_tables()
>   mm: thp: handle split failure in userfaultfd move_pages()
>   mm: thp: handle split failure in device migration
>   mm: huge_mm: Make sure all split_huge_pmd calls are checked
>   mm: thp: allocate PTE page tables lazily at split time
>   mm: thp: remove pgtable_trans_huge_{deposit/withdraw} when not needed
>   mm: thp: add THP_SPLIT_PMD_FAILED counter
>   selftests/mm: add THP PMD split test infrastructure
>   selftests/mm: add partial_mprotect test for change_pmd_range
>   selftests/mm: add partial_mlock test
>   selftests/mm: add partial_mremap test for move_page_tables
>   selftests/mm: add madv_dontneed_partial test
>
>  arch/powerpc/include/asm/book3s/64/pgtable.h  |  12 +-
>  arch/s390/include/asm/pgtable.h               |   6 -
>  arch/s390/mm/pgtable.c                        |  41 ---
>  arch/sparc/include/asm/pgtable_64.h           |   6 -
>  arch/sparc/mm/tlb.c                           |  36 ---
>  include/linux/huge_mm.h                       |  51 +--
>  include/linux/pgtable.h                       |  16 +-
>  include/linux/vm_event_item.h                 |   1 +
>  mm/debug_vm_pgtable.c                         |   4 +-
>  mm/gup.c                                      |  10 +-
>  mm/huge_memory.c                              | 208 +++++++++----
>  mm/khugepaged.c                               |   7 +-
>  mm/memory.c                                   |  26 +-
>  mm/migrate_device.c                           |  33 +-
>  mm/mprotect.c                                 |  11 +-
>  mm/mremap.c                                   |   8 +-
>  mm/pagewalk.c                                 |   8 +-
>  mm/pgtable-generic.c                          |  32 --
>  mm/rmap.c                                     |  42 ++-
>  mm/userfaultfd.c                              |   8 +-
>  mm/vma.c                                      |  37 ++-
>  mm/vmstat.c                                   |   1 +
>  tools/testing/selftests/mm/Makefile           |   1 +
>  .../testing/selftests/mm/thp_pmd_split_test.c | 290 ++++++++++++++++++
>  tools/testing/vma/include/stubs.h             |   9 +-
>  25 files changed, 645 insertions(+), 259 deletions(-)
>  create mode 100644 tools/testing/selftests/mm/thp_pmd_split_test.c
>
> --
> 2.47.3
>