Re: [PATCH v6 14/18] arm64/mm: Implement new [get_and_]clear_full_ptes() batch APIs

[PATCH v6 00/18] Transparent Contiguous PTEs for User Mappings · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
[PATCH v6 01/18] mm: Clarify the spec for set_ptes() · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
[PATCH v6 02/18] mm: thp: Batch-collapse PMD with set_ptes() · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
[PATCH v6 03/18] mm: Introduce pte_advance_pfn() and use for pte_next_pfn() · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 03/18] mm: Introduce pte_advance_pfn() and use for pte_next_pfn() · David Hildenbrand <hidden> · 2024-02-15
[PATCH v6 04/18] arm64/mm: Convert pte_next_pfn() to pte_advance_pfn() · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 04/18] arm64/mm: Convert pte_next_pfn() to pte_advance_pfn() · David Hildenbrand <hidden> · 2024-02-15
Re: [PATCH v6 04/18] arm64/mm: Convert pte_next_pfn() to pte_advance_pfn() · Mark Rutland <mark.rutland@arm.com> · 2024-02-15
Re: [PATCH v6 04/18] arm64/mm: Convert pte_next_pfn() to pte_advance_pfn() · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-15
[PATCH v6 05/18] x86/mm: Convert pte_next_pfn() to pte_advance_pfn() · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 05/18] x86/mm: Convert pte_next_pfn() to pte_advance_pfn() · David Hildenbrand <hidden> · 2024-02-15
[PATCH v6 06/18] mm: Tidy up pte_next_pfn() definition · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 06/18] mm: Tidy up pte_next_pfn() definition · David Hildenbrand <hidden> · 2024-02-15
[PATCH v6 07/18] arm64/mm: Convert READ_ONCE(*ptep) to ptep_get(ptep) · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 07/18] arm64/mm: Convert READ_ONCE(*ptep) to ptep_get(ptep) · Mark Rutland <mark.rutland@arm.com> · 2024-02-15
Re: [PATCH v6 07/18] arm64/mm: Convert READ_ONCE(*ptep) to ptep_get(ptep) · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-15
[PATCH v6 08/18] arm64/mm: Convert set_pte_at() to set_ptes(..., 1) · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 08/18] arm64/mm: Convert set_pte_at() to set_ptes(..., 1) · Mark Rutland <mark.rutland@arm.com> · 2024-02-15
Re: [PATCH v6 08/18] arm64/mm: Convert set_pte_at() to set_ptes(..., 1) · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-15
[PATCH v6 09/18] arm64/mm: Convert ptep_clear() to ptep_get_and_clear() · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 09/18] arm64/mm: Convert ptep_clear() to ptep_get_and_clear() · Mark Rutland <mark.rutland@arm.com> · 2024-02-15
Re: [PATCH v6 09/18] arm64/mm: Convert ptep_clear() to ptep_get_and_clear() · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-15
[PATCH v6 10/18] arm64/mm: New ptep layer to manage contig bit · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 10/18] arm64/mm: New ptep layer to manage contig bit · Mark Rutland <mark.rutland@arm.com> · 2024-02-15
Re: [PATCH v6 10/18] arm64/mm: New ptep layer to manage contig bit · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-15
[PATCH v6 11/18] arm64/mm: Split __flush_tlb_range() to elide trailing DSB · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 11/18] arm64/mm: Split __flush_tlb_range() to elide trailing DSB · Mark Rutland <mark.rutland@arm.com> · 2024-02-15
Re: [PATCH v6 11/18] arm64/mm: Split __flush_tlb_range() to elide trailing DSB · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-15
[PATCH v6 12/18] arm64/mm: Wire up PTE_CONT for user mappings · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 12/18] arm64/mm: Wire up PTE_CONT for user mappings · Mark Rutland <mark.rutland@arm.com> · 2024-02-15
Re: [PATCH v6 12/18] arm64/mm: Wire up PTE_CONT for user mappings · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-16
Re: [PATCH v6 12/18] arm64/mm: Wire up PTE_CONT for user mappings · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-16
Re: [PATCH v6 12/18] arm64/mm: Wire up PTE_CONT for user mappings · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-16
Re: [PATCH v6 12/18] arm64/mm: Wire up PTE_CONT for user mappings · John Hubbard <jhubbard@nvidia.com> · 2024-02-16
Re: [PATCH v6 12/18] arm64/mm: Wire up PTE_CONT for user mappings · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-20
Re: [PATCH v6 12/18] arm64/mm: Wire up PTE_CONT for user mappings · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-19
Re: [PATCH v6 12/18] arm64/mm: Wire up PTE_CONT for user mappings · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-20
[PATCH v6 13/18] arm64/mm: Implement new wrprotect_ptes() batch API · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 13/18] arm64/mm: Implement new wrprotect_ptes() batch API · Mark Rutland <mark.rutland@arm.com> · 2024-02-15
Re: [PATCH v6 13/18] arm64/mm: Implement new wrprotect_ptes() batch API · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-16
[PATCH v6 14/18] arm64/mm: Implement new [get_and_]clear_full_ptes() batch APIs · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 14/18] arm64/mm: Implement new [get_and_]clear_full_ptes() batch APIs · Mark Rutland <mark.rutland@arm.com> · 2024-02-15
Re: [PATCH v6 14/18] arm64/mm: Implement new [get_and_]clear_full_ptes() batch APIs · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-16
[PATCH v6 15/18] mm: Add pte_batch_hint() to reduce scanning in folio_pte_batch() · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
[PATCH v6 16/18] arm64/mm: Implement pte_batch_hint() · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 16/18] arm64/mm: Implement pte_batch_hint() · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-16
[PATCH v6 17/18] arm64/mm: __always_inline to improve fork() perf · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 17/18] arm64/mm: __always_inline to improve fork() perf · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-16
[PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Ryan Roberts <ryan.roberts@arm.com> · 2024-02-15
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Mark Rutland <mark.rutland@arm.com> · 2024-02-15
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Catalin Marinas <catalin.marinas@arm.com> · 2024-02-16
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Kefeng Wang <hidden> · 2024-06-24
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Ryan Roberts <ryan.roberts@arm.com> · 2024-06-24
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Kefeng Wang <hidden> · 2024-06-25
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Baolin Wang <baolin.wang@linux.alibaba.com> · 2024-06-25
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Ryan Roberts <ryan.roberts@arm.com> · 2024-06-25
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Baolin Wang <baolin.wang@linux.alibaba.com> · 2024-06-25
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Ryan Roberts <ryan.roberts@arm.com> · 2024-06-25
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Matthew Wilcox <willy@infradead.org> · 2024-06-25
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Ryan Roberts <ryan.roberts@arm.com> · 2024-06-25
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Matthew Wilcox <willy@infradead.org> · 2024-06-25
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Ryan Roberts <ryan.roberts@arm.com> · 2024-06-25
Re: [PATCH v6 18/18] arm64/mm: Automatically fold contpte mappings · Kefeng Wang <hidden> · 2024-06-25
Re: [PATCH v6 00/18] Transparent Contiguous PTEs for User Mappings · Mark Rutland <mark.rutland@arm.com> · 2024-02-15

From: Catalin Marinas <catalin.marinas@arm.com>
Date: 2024-02-16 12:30:55
Also in: linux-arm-kernel, linux-mm, lkml

On Thu, Feb 15, 2024 at 10:32:01AM +0000, Ryan Roberts wrote:

Optimize the contpte implementation to fix some of the
exit/munmap/dontneed performance regression introduced by the initial
contpte commit. Subsequent patches will solve it entirely.

During exit(), munmap() or madvise(MADV_DONTNEED), mappings must be
cleared. Previously this was done 1 PTE at a time. But the core-mm
supports batched clear via the new [get_and_]clear_full_ptes() APIs. So
let's implement those APIs and for fully covered contpte mappings, we no
longer need to unfold the contpte. This significantly reduces unfolding
operations, reducing the number of tlbis that must be issued.

Tested-by: John Hubbard <jhubbard@nvidia.com>
Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>

Acked-by: Catalin Marinas <catalin.marinas@arm.com>

`h`	back out one level
`j`	next message in thread
`k`	previous message in thread
`l`	drill in
`Esc`	close help / fold thread tree
`?`	toggle this help