[v5,00/17] move pagetable_*_dtor() to __tlb_remove_table()

Message

Qi Zheng Jan. 8, 2025, 6:57 a.m. UTC

Changes in v5:
 - cancel the move of p4d_free_tlb()'s location in [PATCH v4 06/15]
   (Alexander Gordeev)
 - fix the missing pagetable_dtor() in [PATCH v4 08/15] (Kevin Brodsky)
 - change the subject and description in [PATCH v4 12/15]
   (Alexander Gordeev)
 - remove the redundant __HAVE_ARCH_TLB_REMOVE_TABLE definition in [PATCH v4 13/15]
   (Andreas Larsson)
 - add "mm: pgtable: completely move pagetable_dtor() to generic tlb_remove_table()"
   (Kevin Brodsky)
 - add "x86: pgtable: convert __tlb_remove_table() to use struct ptdesc"
 - collect Acked-bys and Reviewed-bys

Changes in v4:
 - remove [PATCH v3 15/17] and [PATCH v3 16/17] (Mike Rapoport)
   (the tlb_remove_page_ptdesc() and tlb_remove_ptdesc() are intermediate
    products of the project: https://kernelnewbies.org/MatthewWilcox/Memdescs,
    so keep them)
 - collect Acked-by

Changes in v3:
 - take patch #5 and #6 from Kevin Brodsky's patch series below.
   Link: https://lore.kernel.org/lkml/[e-mail]/
 - separate the statistics part from [PATCH v2 02/15] as [PATCH v3 04/17], and
   replace the rest part with Kevin Brodsky's patch #6
   (Alexander Gordeev and Kevin Brodsky)
 - change the commit message of [PATCH v2 10/15] and [PATCH v2 11/15]
   (Alexander Gordeev)
 - fix the bug introduced by [PATCH v2 11/15]
   (Peter Zijlstra)
 - rebase onto the next-20241220

Changes in v2:
 - add [PATCH v2 13|14|15/15] (suggested by Peter Zijlstra)
 - add Originally-bys and Suggested-bys
 - rebase onto the next-20241218

Hi all,

As proposed [1] by Peter Zijlstra below, this patch series aims to move
pagetable_*_dtor() into __tlb_remove_table(). This will cleanup pagetable_*_dtor()
a bit and more gracefully fix the UAF issue [2] reported by syzbot.

```
Notably:

 - s390 pud isn't calling the existing pagetable_pud_[cd]tor()
 - none of the p4d things have pagetable_p4d_[cd]tor() (x86,arm64,s390,riscv)
   and they have inconsistent accounting
 - while much of the _ctor calls are in generic code, many of the _dtor
   calls are in arch code for hysterial raisins, this could easily be
   fixed
 - if we fix ptlock_free() to handle NULL, then all the _dtor()
   functions can use it, and we can observe they're all identical
   and can be folded

after all that cleanup, you can move the _dtor from *_free_tlb() into
tlb_remove_table() -- which for the above case, would then have it
called from __tlb_remove_table_free().
```

And hi Andrew, I developed the code based on the latest linux-next, so I reverted
the "mm: pgtable: make ptlock be freed by RCU" first. Once the review of this
patch series is completed, the "mm: pgtable: make ptlock be freed by RCU" can be
dropped directly from mm tree, and this revert patch will not be needed.

This series is based on next-20241220. And I tested this patch series on x86 and
only cross-compiled it on arm, arm64, powerpc, riscv, s390 and sparc.

Comments and suggestions are welcome!

Thanks,
Qi

[1]. https://lore.kernel.org/all/[e-mail]/
[2]. https://lore.kernel.org/all/[e-mail]/

Kevin Brodsky (2):
  riscv: mm: Skip pgtable level check in {pud,p4d}_alloc_one
  asm-generic: pgalloc: Provide generic p4d_{alloc_one,free}

Qi Zheng (15):
  Revert "mm: pgtable: make ptlock be freed by RCU"
  mm: pgtable: add statistics for P4D level page table
  arm64: pgtable: use mmu gather to free p4d level page table
  s390: pgtable: add statistics for PUD and P4D level page table
  mm: pgtable: introduce pagetable_dtor()
  arm: pgtable: move pagetable_dtor() to __tlb_remove_table()
  arm64: pgtable: move pagetable_dtor() to __tlb_remove_table()
  riscv: pgtable: move pagetable_dtor() to __tlb_remove_table()
  x86: pgtable: convert __tlb_remove_table() to use struct ptdesc
  x86: pgtable: move pagetable_dtor() to __tlb_remove_table()
  s390: pgtable: consolidate PxD and PTE TLB free paths
  mm: pgtable: introduce generic __tlb_remove_table()
  mm: pgtable: completely move pagetable_dtor() to generic
    tlb_remove_table()
  mm: pgtable: move __tlb_remove_table_one() in x86 to generic file
  mm: pgtable: introduce generic pagetable_dtor_free()

 Documentation/mm/split_page_table_lock.rst |  4 +-
 arch/arm/include/asm/tlb.h                 | 10 ----
 arch/arm64/include/asm/pgalloc.h           | 18 ------
 arch/arm64/include/asm/tlb.h               | 21 ++++---
 arch/csky/include/asm/pgalloc.h            |  2 +-
 arch/hexagon/include/asm/pgalloc.h         |  2 +-
 arch/loongarch/include/asm/pgalloc.h       |  2 +-
 arch/m68k/include/asm/mcf_pgalloc.h        |  4 +-
 arch/m68k/include/asm/sun3_pgalloc.h       |  2 +-
 arch/m68k/mm/motorola.c                    |  2 +-
 arch/mips/include/asm/pgalloc.h            |  2 +-
 arch/nios2/include/asm/pgalloc.h           |  2 +-
 arch/openrisc/include/asm/pgalloc.h        |  2 +-
 arch/powerpc/include/asm/tlb.h             |  1 +
 arch/powerpc/mm/book3s64/mmu_context.c     |  2 +-
 arch/powerpc/mm/book3s64/pgtable.c         |  2 +-
 arch/powerpc/mm/pgtable-frag.c             |  4 +-
 arch/riscv/include/asm/pgalloc.h           | 69 +++++-----------------
 arch/riscv/include/asm/tlb.h               | 18 ------
 arch/riscv/mm/init.c                       |  4 +-
 arch/s390/include/asm/pgalloc.h            | 31 +++++++---
 arch/s390/include/asm/tlb.h                | 10 ++--
 arch/s390/mm/pgalloc.c                     | 23 +-------
 arch/sh/include/asm/pgalloc.h              |  2 +-
 arch/sparc/include/asm/tlb_64.h            |  1 +
 arch/sparc/mm/init_64.c                    |  2 +-
 arch/sparc/mm/srmmu.c                      |  2 +-
 arch/um/include/asm/pgalloc.h              |  6 +-
 arch/x86/include/asm/pgalloc.h             | 18 ------
 arch/x86/include/asm/tlb.h                 | 33 -----------
 arch/x86/kernel/paravirt.c                 |  5 +-
 arch/x86/mm/pgtable.c                      | 23 ++++----
 include/asm-generic/pgalloc.h              | 55 +++++++++++++++--
 include/asm-generic/tlb.h                  | 24 ++++++--
 include/linux/mm.h                         | 50 ++++++----------
 include/linux/mm_types.h                   |  9 +--
 mm/memory.c                                | 23 +++-----
 mm/mmu_gather.c                            | 20 ++++++-
 38 files changed, 211 insertions(+), 299 deletions(-)