Received-SPF: softfail (google.com: domain of transitioning linux-kernel-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
MIME-Version: 1.0
References: <20220422195516.10769-1-jvgediya@linux.ibm.com> <20220422195516.10769-7-jvgediya@linux.ibm.com>
In-Reply-To: <20220422195516.10769-7-jvgediya@linux.ibm.com>
From:   Wei Xu <weixugc@google.com>
Date:   Fri, 22 Apr 2022 13:47:29 -0700
Message-ID: <CAAPL-u-SDjKGZDmksgvTKyjm=3h53QzU7nRQUfcGtm1noCLnPg@mail.gmail.com>
Subject: Re: [PATCH v3 6/7] mm: demotion: expose per-node demotion targets via sysfs
To:     Jagdish Gediya <jvgediya@linux.ibm.com>
Cc:     linux-mm@kvack.org, linux-kernel@vger.kernel.org,
        akpm@linux-foundation.org, baolin.wang@linux.alibaba.com,
        dave.hansen@linux.intel.com, ying.huang@intel.com,
        aneesh.kumar@linux.ibm.com, shy828301@gmail.com,
        gthelen@google.com, dan.j.williams@intel.com
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Apr 22, 2022 at 12:55 PM Jagdish Gediya <jvgediya@linux.ibm.com> wrote:
>
> Kernel prepares per-node demotion target list based on
> node_states[N_DEMOTION_TARGETS], If enabled through sysfs,
> demotion kicks in during reclaim, and pages get migrated
> according to demotion target list prepared by kernel.
>
> It is helpful to know demotion target list prepared by
> kernel to understand the demotion behaviour, so add
> interface /sys/devices/system/node/nodeX/demotion_targets
> to view per-node demotion targets via sysfs.
>
> Signed-off-by: Jagdish Gediya <jvgediya@linux.ibm.com>
> ---
>  drivers/base/node.c     | 10 ++++++++++
>  include/linux/migrate.h |  1 +
>  mm/migrate.c            | 17 +++++++++++++++++
>  3 files changed, 28 insertions(+)
>
> diff --git a/drivers/base/node.c b/drivers/base/node.c
> index e03eedbc421b..92326219aac2 100644
> --- a/drivers/base/node.c
> +++ b/drivers/base/node.c
> @@ -561,11 +561,21 @@ static ssize_t node_read_distance(struct device *dev,
>  }
>  static DEVICE_ATTR(distance, 0444, node_read_distance, NULL);
>
> +static ssize_t demotion_targets_show(struct device *dev,
> +                                    struct device_attribute *attr, char *buf)
> +{
> +       nodemask_t demotion_targets = node_get_demotion_targets(dev->id);
> +
> +       return sysfs_emit(buf, "%*pbl\n", nodemask_pr_args(&demotion_targets));
> +}
> +static DEVICE_ATTR_RO(demotion_targets);
> +
>  static struct attribute *node_dev_attrs[] = {
>         &dev_attr_meminfo.attr,
>         &dev_attr_numastat.attr,
>         &dev_attr_distance.attr,
>         &dev_attr_vmstat.attr,
> +       &dev_attr_demotion_targets.attr,
>         NULL
>  };
>
> diff --git a/include/linux/migrate.h b/include/linux/migrate.h
> index 90e75d5a54d6..072019441a24 100644
> --- a/include/linux/migrate.h
> +++ b/include/linux/migrate.h
> @@ -173,6 +173,7 @@ int migrate_vma_setup(struct migrate_vma *args);
>  void migrate_vma_pages(struct migrate_vma *migrate);
>  void migrate_vma_finalize(struct migrate_vma *migrate);
>  int next_demotion_node(int node);
> +nodemask_t node_get_demotion_targets(int node);

I think a stub implementation for !CONFIG_MIGRATION is also needed.

>  #else /* CONFIG_MIGRATION disabled: */
>
> diff --git a/mm/migrate.c b/mm/migrate.c
> index 5b92a09fbe4a..da864831bc0c 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -2187,6 +2187,23 @@ struct demotion_nodes {
>
>  static struct demotion_nodes *node_demotion __read_mostly;
>
> +nodemask_t node_get_demotion_targets(int node)
> +{
> +       nodemask_t demotion_targets = NODE_MASK_NONE;
> +       unsigned short target_nr;
> +
> +       if (!node_demotion)
> +               return NODE_MASK_NONE;
> +
> +       rcu_read_lock();
> +       target_nr = READ_ONCE(node_demotion[node].nr);
> +       for (int i = 0; i < target_nr; i++)
> +               node_set(READ_ONCE(node_demotion[node].nodes[i]), demotion_targets);
> +       rcu_read_unlock();
> +
> +       return demotion_targets;
> +}
> +
>  /**
>   * next_demotion_node() - Get the next node in the demotion path
>   * @node: The starting node to lookup the next node
> --
> 2.35.1
>