Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH V3] mm/vmstat: Add events for THP migration without split
To:     John Hubbard <jhubbard@nvidia.com>, linux-mm@kvack.org
Cc:     Daniel Jordan <daniel.m.jordan@oracle.com>,
        Hugh Dickins <hughd@google.com>,
        Matthew Wilcox <willy@infradead.org>, Zi Yan <ziy@nvidia.com>,
        Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        linux-kernel@vger.kernel.org, Jonathan Corbet <corbet@lwn.net>,
        linux-doc@vger.kernel.org
References: <1594080415-27924-1-git-send-email-anshuman.khandual@arm.com>
 <18a3af27-7cab-63d3-29b8-e253021622cc@nvidia.com>
From:   Anshuman Khandual <anshuman.khandual@arm.com>
Message-ID: <f252caee-0af6-8110-b36d-7a92092c34ad@arm.com>
Date:   Wed, 8 Jul 2020 12:44:47 +0530
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.9.1
MIME-Version: 1.0
In-Reply-To: <18a3af27-7cab-63d3-29b8-e253021622cc@nvidia.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On 07/08/2020 05:31 AM, John Hubbard wrote:
> On 2020-07-06 17:06, Anshuman Khandual wrote:
>> Add following new vmstat events which will help in validating THP migration
>> without split. Statistics reported through these new VM events will help in
>> performance debugging.
>>
>> 1. THP_MIGRATION_SUCCESS
>> 2. THP_MIGRATION_FAILURE
>> 3. THP_MIGRATION_SPLIT
>>
>> In addition, these new events also update normal page migration statistics
>> appropriately via PGMIGRATE_SUCCESS and PGMIGRATE_FAILURE. While here, this
>> updates current trace event 'mm_migrate_pages' to accommodate now available
>> THP statistics.
>>
>> Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
>> Cc: Hugh Dickins <hughd@google.com>
>> Cc: Matthew Wilcox <willy@infradead.org>
>> Cc: Zi Yan <ziy@nvidia.com>
>> Cc: John Hubbard <jhubbard@nvidia.com>
>> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
>> Cc: Andrew Morton <akpm@linux-foundation.org>
>> Cc: linux-mm@kvack.org
>> Cc: linux-kernel@vger.kernel.org
>> Signed-off-by: Zi Yan <ziy@nvidia.com>
>> Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
> 
> 
> This is missing the Documentation/ Cc's, did you run scripts/get_maintainer.pl? I'm
> adding linux-doc and Jonathan Corbet, for now at least. And I'll refrain from
> trimming the reply.

I had missed the Documention/ Cc's, thanks for adding them here. Will
update the Cc list next time around.

> 
> 
>> ---
>> Changes in V3:
>>
>> - Formatted new events documentation with 'fmt' tool per Matthew
>> - Made events universally available i.e dropped ARCH_ENABLE_THP_MIGRATION
>> - Added THP_MIGRATION_SPLIT
>> - Updated trace_mm_migrate_pages() with THP events
>> - Made THP events update normal page migration events as well
>>
>> Changes in V2: (https://patchwork.kernel.org/patch/11586893/)
>>
>> - Dropped PMD reference both from code and commit message per Matthew
>> - Added documentation and updated the commit message per Daniel
>>
>> Changes in V1: (https://patchwork.kernel.org/patch/11564497/)
>>
>> - Changed function name as thp_pmd_migration_success() per John
>> - Folded in a fix (https://patchwork.kernel.org/patch/11563009/) from Hugh
>>
>> Changes in RFC V2: (https://patchwork.kernel.org/patch/11554861/)
>>
>> - Decopupled and renamed VM events from their implementation per Zi and John
>> - Added THP_PMD_MIGRATION_FAILURE VM event upon allocation failure and split
>>
>> Changes in RFC V1: (https://patchwork.kernel.org/patch/11542055/)
>>
>>   Documentation/vm/page_migration.rst | 19 +++++++++++
>>   include/linux/vm_event_item.h       |  3 ++
>>   include/trace/events/migrate.h      | 17 ++++++++--
>>   mm/migrate.c                        | 49 ++++++++++++++++++++++++++---
>>   mm/vmstat.c                         |  3 ++
>>   5 files changed, 84 insertions(+), 7 deletions(-)
>>
>> diff --git a/Documentation/vm/page_migration.rst b/Documentation/vm/page_migration.rst
>> index 1d6cd7db4e43..e65d49f3cf86 100644
>> --- a/Documentation/vm/page_migration.rst
>> +++ b/Documentation/vm/page_migration.rst
>> @@ -253,5 +253,24 @@ which are function pointers of struct address_space_operations.
>>        PG_isolated is alias with PG_reclaim flag so driver shouldn't use the flag
>>        for own purpose.
>>   +Quantifying Migration
>> +=====================
>> +Following events can be used to quantify page migration.
>> +
>> +1. PGMIGRATE_SUCCESS       /* Normal page migration success */
> 
> 
> This doesn't explain what's going on with the new combined counter
> behavior. See the proposal below.
> 
> 
>> +2. PGMIGRATE_FAIL          /* Normal page migration failure */
>> +3. THP_MIGRATION_SUCCESS   /* Transparent huge page migration success */
>> +4. THP_MIGRATION_FAILURE   /* Transparent huge page migration failure */
> 
> 
> Shouldn't that be THP_MIGRATION_FAIL, in order to match the existing naming
> scheme here?

Sure, will change it.

> 
> 
>> +5. THP_MIGRATION_SPLIT     /* Transparent huge page got split, retried */
>> +
>> +THP_MIGRATION_SUCCESS is when THP is migrated successfully without getting
>> +split into it's subpages. THP_MIGRATION_FAILURE is when THP could neither
>> +be migrated nor be split. THP_MIGRATION_SPLIT is when THP could not
>> +just be migrated as is but instead get split into it's subpages and later
>> +retried as normal pages. THP events would also update normal page migration
>> +statistics PGMIGRATE_SUCCESS and PGMIGRATE_FAILURE. These events will help
>> +in quantifying and analyzing various THP migration events including both
>> +success and failure cases.
>> +
> 
> 
> I would like to propose the following instead of the above. It's rewritten
> so as to add explanations for the new behavior, remove the redundancy (the last
> sentence in particular is a justification, and belongs in a commit log if anywhere,
> not in the doc itself), fix a few typos, use the full 80 columns, and clarify a
> bit as well. It uses the new THP_MIGRATION_FAIL name:
> 
> 
> Monitoring Migration
> =====================
> 
> The following events (counters) can be used to monitor page migration.
> 
> 1. PGMIGRATE_SUCCESS: Normal page migration success. Each count means that a
>    page was migrated. If the page was a non-THP page, then this counter is
>    increased by one. If the page was a THP, then this counter is increased by
>    the number of THP subpages. For example, migration of a single 2MB THP that
>    has 4KB-size base pages (subpages) will cause this counter to increase by
>    512.
> 
> 2. PGMIGRATE_FAIL: Normal page migration failure. Same counting rules as for
>    _SUCCESS, above: this will be increased by the number of subpages, if it was
>    a THP.
> 
> 3. THP_MIGRATION_SUCCESS: A THP was migrated without being split.
> 
> 4. THP_MIGRATION_FAIL: A THP could not be migrated at all, even after being
>    split.
> 
> 5. THP_MIGRATION_SPLIT: A THP was migrated, but not as such: first, the THP had
>    to be split. After splitting, a migration retry was used for the sub-pages,
>    and that retry succeeded.

The last part might not be true. All that THP_MIGRATION_SPLIT tracks is whether
the THP was split or not. Whether it's subpages got successfully migrated later
or not, would be hidden in overall PGMIGRATE_SUCCESS and PGMIGRATE_FAIL values.

> 
> THP_MIGRATION_* events also update the appropriate PGMIGRATE_SUCCESS or
> PGMIGRATE_FAILURE events. For example, a THP migration failure will cause both
> THP_MIGRATION_FAIL and PGMIGRATE_FAIL to increase.

Above explanation (probably with some changes) looks better, will update. 

> 
> 
>>   Christoph Lameter, May 8, 2006.
>>   Minchan Kim, Mar 28, 2016.
>> diff --git a/include/linux/vm_event_item.h b/include/linux/vm_event_item.h
>> index 24fc7c3ae7d6..5e7ffa025589 100644
>> --- a/include/linux/vm_event_item.h
>> +++ b/include/linux/vm_event_item.h
>> @@ -95,6 +95,9 @@ enum vm_event_item { PGPGIN, PGPGOUT, PSWPIN, PSWPOUT,
>>           THP_ZERO_PAGE_ALLOC_FAILED,
>>           THP_SWPOUT,
>>           THP_SWPOUT_FALLBACK,
>> +        THP_MIGRATION_SUCCESS,
>> +        THP_MIGRATION_FAILURE,
>> +        THP_MIGRATION_SPLIT,
>>   #endif
>>   #ifdef CONFIG_MEMORY_BALLOON
>>           BALLOON_INFLATE,
>> diff --git a/include/trace/events/migrate.h b/include/trace/events/migrate.h
>> index 705b33d1e395..4d434398d64d 100644
>> --- a/include/trace/events/migrate.h
>> +++ b/include/trace/events/migrate.h
>> @@ -46,13 +46,18 @@ MIGRATE_REASON
>>   TRACE_EVENT(mm_migrate_pages,
>>         TP_PROTO(unsigned long succeeded, unsigned long failed,
>> -         enum migrate_mode mode, int reason),
>> +         unsigned long thp_succeeded, unsigned long thp_failed,
>> +         unsigned long thp_split, enum migrate_mode mode, int reason),
>>   -    TP_ARGS(succeeded, failed, mode, reason),
>> +    TP_ARGS(succeeded, failed, thp_succeeded, thp_failed,
>> +        thp_split, mode, reason),
>>         TP_STRUCT__entry(
>>           __field(    unsigned long,        succeeded)
>>           __field(    unsigned long,        failed)
>> +        __field(    unsigned long,        thp_succeeded)
>> +        __field(    unsigned long,        thp_failed)
>> +        __field(    unsigned long,        thp_split)
>>           __field(    enum migrate_mode,    mode)
>>           __field(    int,            reason)
>>       ),
>> @@ -60,13 +65,19 @@ TRACE_EVENT(mm_migrate_pages,
>>       TP_fast_assign(
>>           __entry->succeeded    = succeeded;
>>           __entry->failed        = failed;
>> +        __entry->thp_succeeded    = thp_succeeded;
>> +        __entry->thp_failed    = thp_failed;
>> +        __entry->thp_split    = thp_split;
>>           __entry->mode        = mode;
>>           __entry->reason        = reason;
>>       ),
>>   -    TP_printk("nr_succeeded=%lu nr_failed=%lu mode=%s reason=%s",
>> +    TP_printk("nr_succeeded=%lu nr_failed=%lu nr_thp_succeeded=%lu nr_thp_failed=%lu nr_thp_split=%lu mode=%s reason=%s",
>>           __entry->succeeded,
>>           __entry->failed,
>> +        __entry->thp_succeeded,
>> +        __entry->thp_failed,
>> +        __entry->thp_split,
>>           __print_symbolic(__entry->mode, MIGRATE_MODE),
>>           __print_symbolic(__entry->reason, MIGRATE_REASON))
>>   );
>> diff --git a/mm/migrate.c b/mm/migrate.c
>> index f37729673558..baf3cc477d11 100644
>> --- a/mm/migrate.c
>> +++ b/mm/migrate.c
>> @@ -1429,22 +1429,35 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
>>           enum migrate_mode mode, int reason)
>>   {
>>       int retry = 1;
>> +    int thp_retry = 1;
>>       int nr_failed = 0;
>>       int nr_succeeded = 0;
>> +    int nr_thp_succeeded = 0;
>> +    int nr_thp_failed = 0;
>> +    int nr_thp_split = 0;
>>       int pass = 0;
>> +    bool is_thp = false;
>>       struct page *page;
>>       struct page *page2;
>>       int swapwrite = current->flags & PF_SWAPWRITE;
>> -    int rc;
>> +    int rc, thp_nr_pages;
>>         if (!swapwrite)
>>           current->flags |= PF_SWAPWRITE;
>>   -    for(pass = 0; pass < 10 && retry; pass++) {
>> +    for (pass = 0; pass < 10 && (retry || thp_retry); pass++) {
>>           retry = 0;
>> +        thp_retry = 0;
>>             list_for_each_entry_safe(page, page2, from, lru) {
>>   retry:
>> +            /*
>> +             * THP statistics is based on the source huge page.
>> +             * Capture required information that might get lost
>> +             * during migration.
>> +             */
>> +            is_thp = PageTransHuge(page);
>> +            thp_nr_pages = hpage_nr_pages(page);
>>               cond_resched();
>>                 if (PageHuge(page))
>> @@ -1475,15 +1488,30 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
>>                       unlock_page(page);
>>                       if (!rc) {
>>                           list_safe_reset_next(page, page2, lru);
>> +                        nr_thp_split++;
>>                           goto retry;
>>                       }
>>                   }
>> +                if (is_thp) {
>> +                    nr_thp_failed++;
>> +                    nr_failed += thp_nr_pages;
>> +                    goto out;
>> +                }
>>                   nr_failed++;
>>                   goto out;
>>               case -EAGAIN:
>> +                if (is_thp) {
>> +                    thp_retry++;
>> +                    break;
>> +                }
>>                   retry++;
>>                   break;
>>               case MIGRATEPAGE_SUCCESS:
>> +                if (is_thp) {
>> +                    nr_thp_succeeded++;
>> +                    nr_succeeded += thp_nr_pages;
>> +                    break;
>> +                }
>>                   nr_succeeded++;
>>                   break;
>>               default:
>> @@ -1493,19 +1521,32 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
>>                    * removed from migration page list and not
>>                    * retried in the next outer loop.
>>                    */
>> +                if (is_thp) {
>> +                    nr_thp_failed++;
>> +                    nr_failed += thp_nr_pages;
>> +                    break;
>> +                }
>>                   nr_failed++;
>>                   break;
>>               }
>>           }
>>       }
>> -    nr_failed += retry;
>> +    nr_failed += retry + thp_retry;
>> +    nr_thp_failed += thp_retry;
>>       rc = nr_failed;
>>   out:
>>       if (nr_succeeded)
>>           count_vm_events(PGMIGRATE_SUCCESS, nr_succeeded);
>>       if (nr_failed)
>>           count_vm_events(PGMIGRATE_FAIL, nr_failed);
>> -    trace_mm_migrate_pages(nr_succeeded, nr_failed, mode, reason);
>> +    if (nr_thp_succeeded)
>> +        count_vm_events(THP_MIGRATION_SUCCESS, nr_thp_succeeded);
>> +    if (nr_thp_failed)
>> +        count_vm_events(THP_MIGRATION_FAILURE, nr_thp_failed);
>> +    if (nr_thp_split)
>> +        count_vm_events(THP_MIGRATION_SPLIT, nr_thp_split);
> 
> As mentioned in the other thread, those "ifs" should be removed. I think you
> can also get away with removing the two pre-existing "ifs" in the same
> patch, too, just to keep the whole set consistent and more readable.

Makes sense, will do.