linux-kernel - Re: [RFC V2] mm/vmstat: Add events for PMD based THP migration without split

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Thu, 21 May 2020 09:40:07 +0530
From:   Anshuman Khandual <anshuman.khandual@....com>
To:     HORIGUCHI NAOYA(堀口　直也) 
        <naoya.horiguchi@....com>
Cc:     "linux-mm@...ck.org" <linux-mm@...ck.org>,
        Naoya Horiguchi <n-horiguchi@...jp.nec.com>,
        Zi Yan <ziy@...dia.com>, John Hubbard <jhubbard@...dia.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [RFC V2] mm/vmstat: Add events for PMD based THP migration
 without split



On 05/20/2020 12:45 PM, HORIGUCHI NAOYA(堀口　直也) wrote:
> On Mon, May 18, 2020 at 12:12:36PM +0530, Anshuman Khandual wrote:
>> This adds the following two new VM events which will help in validating PMD
>> based THP migration without split. Statistics reported through these events
>> will help in performance debugging.
>>
>> 1. THP_PMD_MIGRATION_SUCCESS
>> 2. THP_PMD_MIGRATION_FAILURE
>>
>> Cc: Naoya Horiguchi <n-horiguchi@...jp.nec.com>
>> Cc: Zi Yan <ziy@...dia.com>
>> Cc: John Hubbard <jhubbard@...dia.com>
>> Cc: Andrew Morton <akpm@...ux-foundation.org>
>> Cc: linux-mm@...ck.org
>> Cc: linux-kernel@...r.kernel.org
>> Signed-off-by: Anshuman Khandual <anshuman.khandual@....com>
> 
> Hi Anshuman,

Hi Naoya,

> 
> I'm neutral for additinal lines in /proc/vmstat. It's a classic (so widely
> used) but inflexible interface. Users disabling thp are not happy with many
> thp-related lines, but judging from the fact that we already have many

Right, for similar reason, I am not too keen on enabling these counters
without migration being enabled with ARCH_ENABLE_THP_MIGRATION.

> thp-related lines some users really need them. So I feel hard to decide to
> agree or disagree with additional lines.

Currently these are conditional on ARCH_ENABLE_THP_MIGRATION. So we are
not adding these new lines unless it migration is available and enabled.

> 
> I think that tracepoints are the more flexible interfaces for monitoring,
> so I'm interested more in whether thp migration could be monitorable via
> tracepoint. Do you have any idea/plan on it?

Sure, we can add some trace points as well which can give more granular
details regarding THP migration mechanism itself e.g setting and removing
PMD migration entries etc probably with (vaddr, pmdp, pmd) details.

But we will still need /proc/vmstat entries that will be available right
away without requiring additional steps. This simplicity is essential for
folks to consider using these events more often.

Sure, will look into what trace points can be added for THP migration but
in a subsequent patch.

- Anshuman