Re: [RESEND PATCH v3] arm64: enable THP_SWAP for arm64

[Anshuman Khandual <anshuman.khandual@arm.com>]

On 7/19/22 08:58, Huang, Ying wrote:
> Anshuman Khandual <anshuman.khandual@arm.com> writes:
> 
>> On 7/19/22 06:53, Barry Song wrote:
>>> On Tue, Jul 19, 2022 at 12:44 PM Huang, Ying <ying.huang@intel.com> wrote:
>>>>
>>>> Barry Song <21cnbao@gmail.com> writes:
>>>>
>>>>> From: Barry Song <v-songbaohua@oppo.com>
>>>>>
>>>>> THP_SWAP has been proven to improve the swap throughput significantly
>>>>> on x86_64 according to commit bd4c82c22c367e ("mm, THP, swap: delay
>>>>> splitting THP after swapped out").
>>>>> As long as arm64 uses 4K page size, it is quite similar with x86_64
>>>>> by having 2MB PMD THP. THP_SWAP is architecture-independent, thus,
>>>>> enabling it on arm64 will benefit arm64 as well.
>>>>> A corner case is that MTE has an assumption that only base pages
>>>>> can be swapped. We won't enable THP_SWAP for ARM64 hardware with
>>>>> MTE support until MTE is reworked to coexist with THP_SWAP.
>>>>>
>>>>> A micro-benchmark is written to measure thp swapout throughput as
>>>>> below,
>>>>>
>>>>>  unsigned long long tv_to_ms(struct timeval tv)
>>>>>  {
>>>>>       return tv.tv_sec * 1000 + tv.tv_usec / 1000;
>>>>>  }
>>>>>
>>>>>  main()
>>>>>  {
>>>>>       struct timeval tv_b, tv_e;;
>>>>>  #define SIZE 400*1024*1024
>>>>>       volatile void *p = mmap(NULL, SIZE, PROT_READ | PROT_WRITE,
>>>>>                               MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
>>>>>       if (!p) {
>>>>>               perror("fail to get memory");
>>>>>               exit(-1);
>>>>>       }
>>>>>
>>>>>       madvise(p, SIZE, MADV_HUGEPAGE);
>>>>>       memset(p, 0x11, SIZE); /* write to get mem */
>>>>>
>>>>>       gettimeofday(&tv_b, NULL);
>>>>>       madvise(p, SIZE, MADV_PAGEOUT);
>>>>>       gettimeofday(&tv_e, NULL);
>>>>>
>>>>>       printf("swp out bandwidth: %ld bytes/ms\n",
>>>>>                       SIZE/(tv_to_ms(tv_e) - tv_to_ms(tv_b)));
>>>>>  }
>>>>>
>>>>> Testing is done on rk3568 64bit quad core processor Quad Core
>>>>> Cortex-A55 platform - ROCK 3A.
>>>>> thp swp throughput w/o patch: 2734bytes/ms (mean of 10 tests)
>>>>> thp swp throughput w/  patch: 3331bytes/ms (mean of 10 tests)
>>>>>
>>>>> Cc: "Huang, Ying" <ying.huang@intel.com>
>>>>> Cc: Minchan Kim <minchan@kernel.org>
>>>>> Cc: Johannes Weiner <hannes@cmpxchg.org>
>>>>> Cc: Hugh Dickins <hughd@google.com>
>>>>> Cc: Andrea Arcangeli <aarcange@redhat.com>
>>>>> Cc: Anshuman Khandual <anshuman.khandual@arm.com>
>>>>> Cc: Steven Price <steven.price@arm.com>
>>>>> Cc: Yang Shi <shy828301@gmail.com>
>>>>> Signed-off-by: Barry Song <v-songbaohua@oppo.com>
>>>>> ---
>>>>>  -v3:
>>>>>  * refine the commit log;
>>>>>  * add a benchmark result;
>>>>>  * refine the macro of arch_thp_swp_supported
>>>>>  Thanks to the comments of Anshuman, Andrew, Steven
>>>>>
>>>>>  arch/arm64/Kconfig               |  1 +
>>>>>  arch/arm64/include/asm/pgtable.h |  6 ++++++
>>>>>  include/linux/huge_mm.h          | 12 ++++++++++++
>>>>>  mm/swap_slots.c                  |  2 +-
>>>>>  4 files changed, 20 insertions(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
>>>>> index 1652a9800ebe..e1c540e80eec 100644
>>>>> --- a/arch/arm64/Kconfig
>>>>> +++ b/arch/arm64/Kconfig
>>>>> @@ -101,6 +101,7 @@ config ARM64
>>>>>       select ARCH_WANT_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
>>>>>       select ARCH_WANT_LD_ORPHAN_WARN
>>>>>       select ARCH_WANTS_NO_INSTR
>>>>> +     select ARCH_WANTS_THP_SWAP if ARM64_4K_PAGES
>>>>>       select ARCH_HAS_UBSAN_SANITIZE_ALL
>>>>>       select ARM_AMBA
>>>>>       select ARM_ARCH_TIMER
>>>>> diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
>>>>> index 0b6632f18364..78d6f6014bfb 100644
>>>>> --- a/arch/arm64/include/asm/pgtable.h
>>>>> +++ b/arch/arm64/include/asm/pgtable.h
>>>>> @@ -45,6 +45,12 @@
>>>>>       __flush_tlb_range(vma, addr, end, PUD_SIZE, false, 1)
>>>>>  #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
>>>>>
>>>>> +static inline bool arch_thp_swp_supported(void)
>>>>> +{
>>>>> +     return !system_supports_mte();
>>>>> +}
>>>>> +#define arch_thp_swp_supported arch_thp_swp_supported
>>>>> +
>>>>>  /*
>>>>>   * Outside of a few very special situations (e.g. hibernation), we always
>>>>>   * use broadcast TLB invalidation instructions, therefore a spurious page
>>>>> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
>>>>> index de29821231c9..4ddaf6ad73ef 100644
>>>>> --- a/include/linux/huge_mm.h
>>>>> +++ b/include/linux/huge_mm.h
>>>>> @@ -461,4 +461,16 @@ static inline int split_folio_to_list(struct folio *folio,
>>>>>       return split_huge_page_to_list(&folio->page, list);
>>>>>  }
>>>>>
>>>>> +/*
>>>>> + * archs that select ARCH_WANTS_THP_SWAP but don't support THP_SWP due to
>>>>> + * limitations in the implementation like arm64 MTE can override this to
>>>>> + * false
>>>>> + */
>>>>> +#ifndef arch_thp_swp_supported
>>>>> +static inline bool arch_thp_swp_supported(void)
>>>>> +{
>>>>> +     return true;
>>>>> +}
>>>>
>>>> How about the following?
>>>>
>>>> static inline bool arch_wants_thp_swap(void)
>>>> {
>>>>      return IS_ENABLED(ARCH_WANTS_THP_SWAP);
>>>> }
>>>
>>> This looks good. then i'll need to change arm64 to
>>>
>>>  +static inline bool arch_thp_swp_supported(void)
>>>  +{
>>>  +     return IS_ENABLED(ARCH_WANTS_THP_SWAP) &&  !system_supports_mte();
>>>  +}
>>
>> Why ? CONFIG_THP_SWAP depends on ARCH_WANTS_THP_SWAP. In folio_alloc_swap(),
>> IS_ENABLED(CONFIG_THP_SWAP) enabled, will also imply ARCH_WANTS_THP_SWAP too
>> is enabled. Hence checking for ARCH_WANTS_THP_SWAP again does not make sense
>> either in the generic fallback stub, or in arm64 platform override. Because
>> without ARCH_WANTS_THP_SWAP enabled, arch_thp_swp_supported() should never
>> be called in the first place.
> 
> For the only caller now, the checking looks redundant.  But the original
> proposed implementation as follows,
> 
> static inline bool arch_thp_swp_supported(void)
> {
>      return true;
> }
> 
> will return true even on architectures that don't support/want THP swap.

But the function will never be called on for those platforms.

> That will confuse people too.

I dont see how.

> 
> And the "redundant" checking has no run time overhead, because compiler
> will do the trick.
I understand that, but dont think this indirection is necessary.