Re: [PATCH v3 07/26] x86/mm: introduce mm-local region

[Mike Rapoport <rppt@kernel.org>]

On Sun, Jul 26, 2026 at 10:22:40PM +0000, Brendan Jackman wrote:
> Various security features benefit from having process-local address
> mappings within the kernel. Examples include no-direct-map guest_memfd
> [2] and significant optimizations for ASI [1].
> 
> With the currently envisaged usecases, there will be many situations
> where almost no processes have any need for the mm-local region.
> Therefore, avoid its overhead (memory cost of pagetables, alloc/free
> overhead during fork/exit) for processes that don't use it by requiring
> its users to explicitly initialize it via the new mm_local_* API.
> 
> As pointed out by Andy in [0], x86 already has a PGD entry that is local
> to the mm, which is used for the LDT. In a subsequent patch, the LDT
> remap will be unified with the general mm-local region, but to help
> keep the patch to a manageable size, first just introduce the mm-local
> region.
> 
> On 64-bit, give the mm-local region a whole PGD. On 32-bit, just give it
> one PMD. No investigation has been done into whether it's feasible to
> expand the region on 32-bit. Most likely there is no strong usecase for
> that anyway.
> 
> In order to combine the need for an on-demand mm initialisation, with
> the desire to transparently handle propagating mappings to userspace
> under KPTI, the user and kernel pagetables are shared at the highest
> level possible. For PAE that means the PTE table is shared and for
> 64-bit the P4D/PUD. This is implemented by pre-allocating the first
> shared table when the mm-local region is first initialised.
> 
> [0] https://lore.kernel.org/linux-mm/CALCETrXHbS9VXfZ80kOjiTrreM2EbapYeGp68mvJPbosUtorYA@mail.gmail.com/
> [1] https://linuxasi.dev/
> [2] https://lore.kernel.org/all/20250924151101.2225820-1-patrick.roy@campus.lmu.de
> Signed-off-by: Brendan Jackman <jackmanb@google.com>
> ---
>  arch/x86/Kconfig                        |   2 +
>  arch/x86/include/asm/mmu_context.h      | 118 +++++++++++++++++++++++++++++++-
>  arch/x86/include/asm/pgtable_32_areas.h |   9 ++-
>  arch/x86/include/asm/pgtable_64_types.h |  12 +++-
>  arch/x86/mm/pgtable.c                   |   3 +
>  include/linux/mm.h                      |  13 ++++
>  include/linux/mm_types.h                |   2 +
>  kernel/fork.c                           |   1 +
>  mm/Kconfig                              |   7 ++
>  9 files changed, 161 insertions(+), 6 deletions(-)
> 
> diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> index fb298e2191792..3efab3524a6cf 100644
> --- a/arch/x86/Kconfig
> +++ b/arch/x86/Kconfig
> @@ -132,6 +132,8 @@ config X86
>  	select ARCH_SUPPORTS_LTO_CLANG
>  	select ARCH_SUPPORTS_LTO_CLANG_THIN
>  	select ARCH_SUPPORTS_RT
> +	# LDT remap temporarily clashes with mm-local region, can't have both.
> +	select ARCH_SUPPORTS_MM_LOCAL_REGION	if X86_64 || X86_PAE && !MODIFY_LDT_SYSCALL
>  	select ARCH_USE_BUILTIN_BSWAP
>  	select ARCH_USE_CMPXCHG_LOCKREF
>  	select ARCH_USE_MEMTEST
> diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h
> index ef5b507de34e2..3d4f54673014f 100644
> --- a/arch/x86/include/asm/mmu_context.h
> +++ b/arch/x86/include/asm/mmu_context.h
> @@ -8,8 +8,10 @@
>  
>  #include <trace/events/tlb.h>
>  
> +#include <asm/tlb.h>
>  #include <asm/tlbflush.h>
>  #include <asm/paravirt.h>
> +#include <asm/pgalloc.h>
>  #include <asm/debugreg.h>
>  #include <asm/gsseg.h>
>  #include <asm/desc.h>
> @@ -223,10 +225,124 @@ static inline int arch_dup_mmap(struct mm_struct *oldmm, struct mm_struct *mm)
>  	return ldt_dup_context(oldmm, mm);
>  }
>  
> +#ifdef CONFIG_MM_LOCAL_REGION
> +static inline void mm_local_region_free(struct mm_struct *mm)
> +{
> +	if (!mm_local_region_used(mm))
> +		return;
> +
> +	struct mmu_gather tlb;
> +	unsigned long start = MM_LOCAL_BASE_ADDR;
> +	unsigned long end = MM_LOCAL_END_ADDR;
> +
> +	/*
> +	 * Although free_pgd_range() is intended for freeing user
> +	 * page-tables, it also works out for kernel mappings on x86.
> +	 * Use tlb_gather_mmu_fullmm() to avoid confusing the
> +	 * range-tracking logic in __tlb_adjust_range().
> +	 */
> +	tlb_gather_mmu_fullmm(&tlb, mm);
> +	free_pgd_range(&tlb, start, end, start, end);
> +	tlb_finish_mmu(&tlb);
> +
> +	mm_flags_clear(MMF_LOCAL_REGION_USED, mm);
> +}
> +
> +#if defined(CONFIG_MITIGATION_PAGE_TABLE_ISOLATION) && defined(CONFIG_X86_PAE)
> +static inline pmd_t *pgd_to_pmd_walk(pgd_t *pgd, unsigned long va)

There's very similar mm_find_pmd() in mm/rmap.c and I bet a bunch of other
places walk from PGD to PMD and return PMD in the end.

Can we put this function into, say, mm/pgtable-generic.c?
Finding all the places that do such walk and sticking it there should not
be a part of this set IMNHO, but having it in the generic code is a good
start for a future cleanup.

> +{
> +	p4d_t *p4d;
> +	pud_t *pud;
> +
> +	if (pgd->pgd == 0)
> +		return NULL;
> +
> +	p4d = p4d_offset(pgd, va);
> +	if (p4d_none(*p4d))
> +		return NULL;
> +
> +	pud = pud_offset(p4d, va);
> +	if (pud_none(*pud))
> +		return NULL;
> +
> +	return pmd_offset(pud, va);
> +}

-- 
Sincerely yours,
Mike.