Re: [PATCH v5 1/4] mm: add per-order mTHP anon_fault_alloc and anon_fault_fallback counters

[Ryan Roberts <ryan.roberts@arm.com>]

On 12/04/2024 11:17, Barry Song wrote:
> On Fri, Apr 12, 2024 at 9:56 PM Ryan Roberts <ryan.roberts@arm.com> wrote:
>>
>> On 12/04/2024 10:43, Barry Song wrote:
>>> On Fri, Apr 12, 2024 at 9:27 PM Ryan Roberts <ryan.roberts@arm.com> wrote:
>>>>
>>>> Hi Barry,
>>>>
>>>> 2 remaining comments - otherwise looks good. (same comments I just made in the
>>>> v4 conversation).
>>>>
>>>> On 12/04/2024 08:37, Barry Song wrote:
>>>>> From: Barry Song <v-songbaohua@oppo.com>
>>>>>
>>>>> Profiling a system blindly with mTHP has become challenging due to the
>>>>> lack of visibility into its operations.  Presenting the success rate of
>>>>> mTHP allocations appears to be pressing need.
>>>>>
>>>>> Recently, I've been experiencing significant difficulty debugging
>>>>> performance improvements and regressions without these figures.  It's
>>>>> crucial for us to understand the true effectiveness of mTHP in real-world
>>>>> scenarios, especially in systems with fragmented memory.
>>>>>
>>>>> This patch establishes the framework for per-order mTHP
>>>>> counters. It begins by introducing the anon_fault_alloc and
>>>>> anon_fault_fallback counters. Additionally, to maintain consistency
>>>>> with thp_fault_fallback_charge in /proc/vmstat, this patch also tracks
>>>>> anon_fault_fallback_charge when mem_cgroup_charge fails for mTHP.
>>>>> Incorporating additional counters should now be straightforward as well.
>>>>>
>>>>> Signed-off-by: Barry Song <v-songbaohua@oppo.com>
>>>>> Cc: Chris Li <chrisl@kernel.org>
>>>>> Cc: David Hildenbrand <david@redhat.com>
>>>>> Cc: Domenico Cerasuolo <cerasuolodomenico@gmail.com>
>>>>> Cc: Kairui Song <kasong@tencent.com>
>>>>> Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
>>>>> Cc: Peter Xu <peterx@redhat.com>
>>>>> Cc: Ryan Roberts <ryan.roberts@arm.com>
>>>>> Cc: Suren Baghdasaryan <surenb@google.com>
>>>>> Cc: Yosry Ahmed <yosryahmed@google.com>
>>>>> Cc: Yu Zhao <yuzhao@google.com>
>>>>> ---
>>>>>  include/linux/huge_mm.h | 51 ++++++++++++++++++++++++++++++++++
>>>>>  mm/huge_memory.c        | 61 +++++++++++++++++++++++++++++++++++++++++
>>>>>  mm/memory.c             |  3 ++
>>>>>  mm/page_alloc.c         |  4 +++
>>>>>  4 files changed, 119 insertions(+)
>>>>>
>>>>> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
>>>>> index e896ca4760f6..c5beb54b97cb 100644
>>>>> --- a/include/linux/huge_mm.h
>>>>> +++ b/include/linux/huge_mm.h
>>>>> @@ -264,6 +264,57 @@ unsigned long thp_vma_allowable_orders(struct vm_area_struct *vma,
>>>>>                                         enforce_sysfs, orders);
>>>>>  }
>>>>>
>>>>> +enum mthp_stat_item {
>>>>> +     MTHP_STAT_ANON_FAULT_ALLOC,
>>>>> +     MTHP_STAT_ANON_FAULT_FALLBACK,
>>>>> +     MTHP_STAT_ANON_FAULT_FALLBACK_CHARGE,
>>>>> +     __MTHP_STAT_COUNT
>>>>> +};
>>>>> +
>>>>> +struct mthp_stat {
>>>>> +     unsigned long stats[0][__MTHP_STAT_COUNT];
>>>>> +};
>>>>> +
>>>>> +extern struct mthp_stat __percpu *mthp_stats;
>>>>> +
>>>>> +static inline void count_mthp_stat(int order, enum mthp_stat_item item)
>>>>> +{
>>>>> +     if (order <= 0 || order > PMD_ORDER || !mthp_stats)
>>>>> +             return;
>>>>> +
>>>>> +     this_cpu_inc(mthp_stats->stats[order][item]);
>>>>> +}
>>>>> +
>>>>> +static inline void count_mthp_stats(int order, enum mthp_stat_item item, long delta)
>>>>> +{
>>>>> +     if (order <= 0 || order > PMD_ORDER || !mthp_stats)
>>>>> +             return;
>>>>> +
>>>>> +     this_cpu_add(mthp_stats->stats[order][item], delta);
>>>>> +}
>>>>> +
>>>>> +/*
>>>>> + * Fold the foreign cpu mthp stats into our own.
>>>>> + *
>>>>> + * This is adding to the stats on one processor
>>>>> + * but keeps the global counts constant.
>>>>> + */
>>>>> +static inline void mthp_stats_fold_cpu(int cpu)
>>>>> +{
>>>>> +     struct mthp_stat *fold_stat;
>>>>> +     int i, j;
>>>>> +
>>>>> +     if (!mthp_stats)
>>>>> +             return;
>>>>> +     fold_stat = per_cpu_ptr(mthp_stats, cpu);
>>>>> +     for (i = 1; i <= PMD_ORDER; i++) {
>>>>> +             for (j = 0; j < __MTHP_STAT_COUNT; j++) {
>>>>> +                     count_mthp_stats(i, j, fold_stat->stats[i][j]);
>>>>> +                     fold_stat->stats[i][j] = 0;
>>>>> +             }
>>>>> +     }
>>>>> +}
>>>>
>>>> This is a pretty horrible hack; I'm pretty sure just summing for all *possible*
>>>> cpus should work.
>>>>
>>>>> +
>>>>>  #define transparent_hugepage_use_zero_page()                         \
>>>>>       (transparent_hugepage_flags &                                   \
>>>>>        (1<<TRANSPARENT_HUGEPAGE_USE_ZERO_PAGE_FLAG))
>>>>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
>>>>> index dc30139590e6..21c4ac74b484 100644
>>>>> --- a/mm/huge_memory.c
>>>>> +++ b/mm/huge_memory.c
>>>>> @@ -526,6 +526,50 @@ static const struct kobj_type thpsize_ktype = {
>>>>>       .sysfs_ops = &kobj_sysfs_ops,
>>>>>  };
>>>>>
>>>>> +struct mthp_stat __percpu *mthp_stats;
>>>>> +
>>>>> +static unsigned long sum_mthp_stat(int order, enum mthp_stat_item item)
>>>>> +{
>>>>> +     unsigned long sum = 0;
>>>>> +     int cpu;
>>>>> +
>>>>> +     cpus_read_lock();
>>>>> +     for_each_online_cpu(cpu) {
>>>>> +             struct mthp_stat *this = per_cpu_ptr(mthp_stats, cpu);
>>>>> +
>>>>> +             sum += this->stats[order][item];
>>>>> +     }
>>>>> +     cpus_read_unlock();
>>>>> +
>>>>> +     return sum;
>>>>> +}
>>>>> +
>>>>> +#define DEFINE_MTHP_STAT_ATTR(_name, _index)                                 \
>>>>> +static ssize_t _name##_show(struct kobject *kobj,                    \
>>>>> +                     struct kobj_attribute *attr, char *buf)         \
>>>>> +{                                                                    \
>>>>> +     int order = to_thpsize(kobj)->order;                            \
>>>>> +                                                                     \
>>>>> +     return sysfs_emit(buf, "%lu\n", sum_mthp_stat(order, _index));  \
>>>>> +}                                                                    \
>>>>> +static struct kobj_attribute _name##_attr = __ATTR_RO(_name)
>>>>> +
>>>>> +DEFINE_MTHP_STAT_ATTR(anon_fault_alloc, MTHP_STAT_ANON_FAULT_ALLOC);
>>>>> +DEFINE_MTHP_STAT_ATTR(anon_fault_fallback, MTHP_STAT_ANON_FAULT_FALLBACK);
>>>>> +DEFINE_MTHP_STAT_ATTR(anon_fault_fallback_charge, MTHP_STAT_ANON_FAULT_FALLBACK_CHARGE);
>>>>> +
>>>>> +static struct attribute *stats_attrs[] = {
>>>>> +     &anon_fault_alloc_attr.attr,
>>>>> +     &anon_fault_fallback_attr.attr,
>>>>> +     &anon_fault_fallback_charge_attr.attr,
>>>>> +     NULL,
>>>>> +};
>>>>> +
>>>>> +static struct attribute_group stats_attr_group = {
>>>>> +     .name = "stats",
>>>>> +     .attrs = stats_attrs,
>>>>> +};
>>>>> +
>>>>>  static struct thpsize *thpsize_create(int order, struct kobject *parent)
>>>>>  {
>>>>>       unsigned long size = (PAGE_SIZE << order) / SZ_1K;
>>>>> @@ -549,6 +593,12 @@ static struct thpsize *thpsize_create(int order, struct kobject *parent)
>>>>>               return ERR_PTR(ret);
>>>>>       }
>>>>>
>>>>> +     ret = sysfs_create_group(&thpsize->kobj, &stats_attr_group);
>>>>> +     if (ret) {
>>>>> +             kobject_put(&thpsize->kobj);
>>>>> +             return ERR_PTR(ret);
>>>>> +     }
>>>>> +
>>>>>       thpsize->order = order;
>>>>>       return thpsize;
>>>>>  }
>>>>> @@ -691,6 +741,11 @@ static int __init hugepage_init(void)
>>>>>        */
>>>>>       MAYBE_BUILD_BUG_ON(HPAGE_PMD_ORDER < 2);
>>>>>
>>>>> +     mthp_stats = __alloc_percpu((PMD_ORDER + 1) * sizeof(mthp_stats->stats[0]),
>>>>> +                     sizeof(unsigned long));
>>>>
>>>> Personally I think it would be cleaner to allocate statically using
>>>> ilog2(MAX_PTRS_PER_PTE) instead of PMD_ORDER.
>>>
>>> Hi Ryan,
>>>
>>> I don't understand why MAX_PTRS_PER_PTE is the correct size. For ARM64,
>>>
>>> #define PMD_ORDER       (PMD_SHIFT - PAGE_SHIFT)
>>>
>>> #define MAX_PTRS_PER_PTE PTRS_PER_PTE
>>>
>>> #define PTRS_PER_PTE            (1 << (PAGE_SHIFT - 3))
>>>
>>> while PAGE_SIZE is 16KiB or 64KiB, PTRS_PER_PTE can be a huge number?
>>>
>>>
>>> Am I missing something?
>>
>> PTRS_PER_PTE is the number of PTE entries in a PTE table. On arm64 its as follows:
>>
>> PAGE_SIZE       PAGE_SHIFT      PTRS_PER_PTE
>> 4K              12              512
>> 16K             14              2048
>> 64K             16              8192
>>
>> So (PTRS_PER_PTE * PAGE_SIZE) = PMD_SIZE
>>
>> PMD_ORDER is ilog2(PMD_SIZE / PAGE_SIZE) = ilog2(PTRS_PER_PTE)
>>
>> MAX_PTRS_PER_PTE is just the maximum value that PTRS_PER_PTE will ever have,
>> (and its equal to PTRS_PER_PTE except for powerpc).
>>
>> Pretty sure the math is correct?
> 
> I am not convinced the math is correct :-)
> 
> while page size is 64KiB, the page table is as below,
> PMD_ORDER = L2 index bits = [41:29] = 13 != ilog2(8192)

1 << 13 = 8192

Right? So:

ilog2(8192) = 13

What's wrong with that?

I even checked in Python to make sure I'm not going mad:

>>> import math
>>> math.log2(8192)
13.0

> 
> 
> +--------+--------+--------+--------+--------+--------+--------+--------+
> |63    56|55    48|47    40|39    32|31    24|23    16|15     8|7      0|
> +--------+--------+--------+--------+--------+--------+--------+--------+
>  |                 |    |               |              |
>  |                 |    |               |              v
>  |                 |    |               |            [15:0]  in-page offset
>  |                 |    |               +----------> [28:16] L3 index
>  |                 |    +--------------------------> [41:29] L2 index
>  |                 +-------------------------------> [47:42] L1 index (48-bit)
>  |                                                   [51:42] L1 index (52-bit)
>  +-------------------------------------------------> [63] TTBR0/1
>  
> while page size is 4KiB, the page table is as below,
> 
> +--------+--------+--------+--------+--------+--------+--------+--------+
> |63    56|55    48|47    40|39    32|31    24|23    16|15     8|7      0|
> +--------+--------+--------+--------+--------+--------+--------+--------+
>  |                 |         |         |         |         |
>  |                 |         |         |         |         v
>  |                 |         |         |         |   [11:0]  in-page offset
>  |                 |         |         |         +-> [20:12] L3 index
>  |                 |         |         +-----------> [29:21] L2 index
>  |                 |         +---------------------> [38:30] L1 index
>  |                 +-------------------------------> [47:39] L0 index
>  +-------------------------------------------------> [63] TTBR0/1
> 
> PMD_ORDER = L2 index bits = [29:21] = 9 = ilog2(512).
> 
> You are only correct while page size = 4KiB.
> 
> 
> 
>