Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220830081736.119281-1-aneesh.kumar@linux.ibm.com>
 <87tu5rzigc.fsf@yhuang6-desk2.ccr.corp.intel.com> <ad19e105-9290-922d-59e6-e6813a30f5f0@linux.ibm.com>
 <87pmgezkhp.fsf@yhuang6-desk2.ccr.corp.intel.com>
In-Reply-To: <87pmgezkhp.fsf@yhuang6-desk2.ccr.corp.intel.com>
From:   Wei Xu <weixugc@google.com>
Date:   Thu, 1 Sep 2022 22:09:13 -0700
Message-ID: <CAAPL-u8MEs04DkHy6kaS788VjdYZZjAYOgzMnioOzDXbc0ZhhQ@mail.gmail.com>
Subject: Re: [PATCH v3 updated] mm/demotion: Expose memory tier details via sysfs
To:     "Huang, Ying" <ying.huang@intel.com>
Cc:     Aneesh Kumar K V <aneesh.kumar@linux.ibm.com>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Linux MM <linux-mm@kvack.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Yang Shi <shy828301@gmail.com>,
        Davidlohr Bueso <dave@stgolabs.net>,
        Tim C Chen <tim.c.chen@intel.com>,
        Michal Hocko <mhocko@kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Hesham Almatary <hesham.almatary@huawei.com>,
        Dave Hansen <dave.hansen@intel.com>,
        Jonathan Cameron <Jonathan.Cameron@huawei.com>,
        Alistair Popple <apopple@nvidia.com>,
        Dan Williams <dan.j.williams@intel.com>,
        jvgediya.oss@gmail.com, Bharata B Rao <bharata@amd.com>,
        Greg Thelen <gthelen@google.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Sep 1, 2022 at 5:33 PM Huang, Ying <ying.huang@intel.com> wrote:
>
> Aneesh Kumar K V <aneesh.kumar@linux.ibm.com> writes:
>
> > On 9/1/22 12:31 PM, Huang, Ying wrote:
> >> "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com> writes:
> >>
> >>> This patch adds /sys/devices/virtual/memory_tiering/ where all memory=
 tier
> >>> related details can be found. All allocated memory tiers will be list=
ed
> >>> there as /sys/devices/virtual/memory_tiering/memory_tierN/
> >>>
> >>> The nodes which are part of a specific memory tier can be listed via
> >>> /sys/devices/virtual/memory_tiering/memory_tierN/nodes
> >>
> >> I think "memory_tier" is a better subsystem/bus name than
> >> memory_tiering.  Because we have a set of memory_tierN devices inside.
> >> "memory_tier" sounds more natural.  I know this is subjective, just my
> >> preference.
> >>
> >>>
> >>> A directory hierarchy looks like
> >>> :/sys/devices/virtual/memory_tiering$ tree memory_tier4/
> >>> memory_tier4/
> >>> =E2=94=9C=E2=94=80=E2=94=80 nodes
> >>> =E2=94=9C=E2=94=80=E2=94=80 subsystem -> ../../../../bus/memory_tieri=
ng
> >>> =E2=94=94=E2=94=80=E2=94=80 uevent
> >>>
> >>> All toptier nodes are listed via
> >>> /sys/devices/virtual/memory_tiering/toptier_nodes
> >>>
> >>> :/sys/devices/virtual/memory_tiering$ cat toptier_nodes
> >>> 0,2
> >>> :/sys/devices/virtual/memory_tiering$ cat memory_tier4/nodes
> >>> 0,2
> >>
> >> I don't think that it is a good idea to show toptier information in us=
er
> >> space interface.  Because it is just a in kernel implementation
> >> details.  Now, we only promote pages from !toptier to toptier.  But
> >> there may be multiple memory tiers in toptier and !toptier, we may
> >> change the implementation in the future.  For example, we may promote
> >> pages from DRAM to HBM in the future.
> >>
> >
> >
> > In the case you describe above and others, we will always have a list o=
f
> > NUMA nodes from which memory promotion is not done.
> > /sys/devices/virtual/memory_tiering/toptier_nodes shows that list.
>
> I don't think we will need that interface if we don't restrict promotion
> in the future.  For example, he can just check the memory tier with
> smallest number.
>
> TBH, I don't know why do we need that interface.  What is it for?  We
> don't want to expose unnecessary information to restrict our in kernel
> implementation in the future.
>
> So, please remove that interface at least before we discussing it
> thoroughly.

I have asked for this interface to allow the userspace to query a list
of top-tier nodes as the targets of userspace-driven promotions.  The
idea is that demotion can gradually go down tier by tier, but we
promote hot pages directly to the top-tier and bypass the immediate
tiers.

Certainly, this can be viewed as a policy choice.  Given that now we
have a clearly defined memory tier hierarchy in sysfs and the
toptier_nodes content can be constructed from this memory tier
hierarchy and other information from the node sysfs interfaces, I am
fine if we want to remove toptier_nodes and keep the current memory
tier sysfs interfaces to the minimal.

Wei Xu

> >> Do we need a way to show the default memory tier in sysfs?  That is, t=
he
> >> memory tier that the DRAM nodes belong to.
> >>
> >
> > I will hold adding that until we have support for modifying memory tier=
 details from
> > userspace. That is when userspace would want to know about the default =
memory tier.
> >
> > For now, the user interface is a simpler hierarchy of memory tiers, it'=
s associated
> > nodes and the list of nodes from which promotion is not done.
>
> OK.
>
> Best Regards,
> Huang, Ying