linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Zhu Yanhai <zhu.yanhai@gmail.com>
To: Ying Han <yinghan@google.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>,
	Pavel Emelyanov <xemul@openvz.org>,
	Balbir Singh <balbir@linux.vnet.ibm.com>,
	Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>,
	Li Zefan <lizf@cn.fujitsu.com>, Mel Gorman <mel@csn.ul.ie>,
	Christoph Lameter <cl@linux.com>,
	Johannes Weiner <hannes@cmpxchg.org>,
	Rik van Riel <riel@redhat.com>, Hugh Dickins <hughd@google.com>,
	KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>,
	Tejun Heo <tj@kernel.org>, Michal Hocko <mhocko@suse.cz>,
	Andrew Morton <akpm@linux-foundation.org>,
	Dave Hansen <dave@linux.vnet.ibm.com>,
	linux-mm@kvack.org
Subject: Re: [PATCH V3 4/7] Infrastructure to support per-memcg reclaim.
Date: Thu, 14 Apr 2011 11:57:55 +0800	[thread overview]
Message-ID: <BANLkTi=Qdv3RZuhN++OX2-S8OZqrL7=KBg@mail.gmail.com> (raw)
In-Reply-To: <1302678187-24154-5-git-send-email-yinghan@google.com>

Hi Ying,

2011/4/13 Ying Han <yinghan@google.com>:
> -extern int kswapd_run(int nid);
> -extern void kswapd_stop(int nid);
> +extern int kswapd_run(int nid, struct mem_cgroup *mem);
> +extern void kswapd_stop(int nid, struct mem_cgroup *mem);

This breaks online_pages() and offline_pages(), which are also
the caller of kswaped_run() and kswaped_stop().

Thanks,
Zhu Yanhai

>
>  #ifdef CONFIG_MMU
>  /* linux/mm/shmem.c */
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index 36ae377..acd84a8 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -274,6 +274,8 @@ struct mem_cgroup {
>        spinlock_t pcp_counter_lock;
>
>        int wmark_ratio;
> +
> +       wait_queue_head_t *kswapd_wait;
>  };
>
>  /* Stuffs for move charges at task migration. */
> @@ -4622,6 +4624,33 @@ int mem_cgroup_watermark_ok(struct mem_cgroup *mem,
>        return ret;
>  }
>
> +int mem_cgroup_init_kswapd(struct mem_cgroup *mem, struct kswapd *kswapd_p)
> +{
> +       if (!mem || !kswapd_p)
> +               return 0;
> +
> +       mem->kswapd_wait = &kswapd_p->kswapd_wait;
> +       kswapd_p->kswapd_mem = mem;
> +
> +       return css_id(&mem->css);
> +}
> +
> +void mem_cgroup_clear_kswapd(struct mem_cgroup *mem)
> +{
> +       if (mem)
> +               mem->kswapd_wait = NULL;
> +
> +       return;
> +}
> +
> +wait_queue_head_t *mem_cgroup_kswapd_wait(struct mem_cgroup *mem)
> +{
> +       if (!mem)
> +               return NULL;
> +
> +       return mem->kswapd_wait;
> +}
> +
>  static int mem_cgroup_soft_limit_tree_init(void)
>  {
>        struct mem_cgroup_tree_per_node *rtpn;
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index 77ac74f..a1a1211 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -2242,6 +2242,7 @@ static bool pgdat_balanced(pg_data_t *pgdat, unsigned long balanced_pages,
>  }
>
>  static DEFINE_SPINLOCK(kswapds_spinlock);
> +#define is_node_kswapd(kswapd_p) (!(kswapd_p)->kswapd_mem)
>
>  /* is kswapd sleeping prematurely? */
>  static int sleeping_prematurely(struct kswapd *kswapd, int order,
> @@ -2251,11 +2252,16 @@ static int sleeping_prematurely(struct kswapd *kswapd, int order,
>        unsigned long balanced = 0;
>        bool all_zones_ok = true;
>        pg_data_t *pgdat = kswapd->kswapd_pgdat;
> +       struct mem_cgroup *mem = kswapd->kswapd_mem;
>
>        /* If a direct reclaimer woke kswapd within HZ/10, it's premature */
>        if (remaining)
>                return true;
>
> +       /* Doesn't support for per-memcg reclaim */
> +       if (mem)
> +               return false;
> +
>        /* Check the watermark levels */
>        for (i = 0; i < pgdat->nr_zones; i++) {
>                struct zone *zone = pgdat->node_zones + i;
> @@ -2598,19 +2604,25 @@ static void kswapd_try_to_sleep(struct kswapd *kswapd_p, int order,
>         * go fully to sleep until explicitly woken up.
>         */
>        if (!sleeping_prematurely(kswapd_p, order, remaining, classzone_idx)) {
> -               trace_mm_vmscan_kswapd_sleep(pgdat->node_id);
> +               if (is_node_kswapd(kswapd_p)) {
> +                       trace_mm_vmscan_kswapd_sleep(pgdat->node_id);
>
> -               /*
> -                * vmstat counters are not perfectly accurate and the estimated
> -                * value for counters such as NR_FREE_PAGES can deviate from the
> -                * true value by nr_online_cpus * threshold. To avoid the zone
> -                * watermarks being breached while under pressure, we reduce the
> -                * per-cpu vmstat threshold while kswapd is awake and restore
> -                * them before going back to sleep.
> -                */
> -               set_pgdat_percpu_threshold(pgdat, calculate_normal_threshold);
> -               schedule();
> -               set_pgdat_percpu_threshold(pgdat, calculate_pressure_threshold);
> +                       /*
> +                        * vmstat counters are not perfectly accurate and the
> +                        * estimated value for counters such as NR_FREE_PAGES
> +                        * can deviate from the true value by nr_online_cpus *
> +                        * threshold. To avoid the zone watermarks being
> +                        * breached while under pressure, we reduce the per-cpu
> +                        * vmstat threshold while kswapd is awake and restore
> +                        * them before going back to sleep.
> +                        */
> +                       set_pgdat_percpu_threshold(pgdat,
> +                                                  calculate_normal_threshold);
> +                       schedule();
> +                       set_pgdat_percpu_threshold(pgdat,
> +                                               calculate_pressure_threshold);
> +               } else
> +                       schedule();
>        } else {
>                if (remaining)
>                        count_vm_event(KSWAPD_LOW_WMARK_HIT_QUICKLY);
> @@ -2620,6 +2632,12 @@ static void kswapd_try_to_sleep(struct kswapd *kswapd_p, int order,
>        finish_wait(wait_h, &wait);
>  }
>
> +static unsigned long balance_mem_cgroup_pgdat(struct mem_cgroup *mem_cont,
> +                                                       int order)
> +{
> +       return 0;
> +}
> +
>  /*
>  * The background pageout daemon, started as a kernel thread
>  * from the init process.
> @@ -2639,6 +2657,7 @@ int kswapd(void *p)
>        int classzone_idx;
>        struct kswapd *kswapd_p = (struct kswapd *)p;
>        pg_data_t *pgdat = kswapd_p->kswapd_pgdat;
> +       struct mem_cgroup *mem = kswapd_p->kswapd_mem;
>        wait_queue_head_t *wait_h = &kswapd_p->kswapd_wait;
>        struct task_struct *tsk = current;
>
> @@ -2649,10 +2668,12 @@ int kswapd(void *p)
>
>        lockdep_set_current_reclaim_state(GFP_KERNEL);
>
> -       BUG_ON(pgdat->kswapd_wait != wait_h);
> -       cpumask = cpumask_of_node(pgdat->node_id);
> -       if (!cpumask_empty(cpumask))
> -               set_cpus_allowed_ptr(tsk, cpumask);
> +       if (is_node_kswapd(kswapd_p)) {
> +               BUG_ON(pgdat->kswapd_wait != wait_h);
> +               cpumask = cpumask_of_node(pgdat->node_id);
> +               if (!cpumask_empty(cpumask))
> +                       set_cpus_allowed_ptr(tsk, cpumask);
> +       }
>        current->reclaim_state = &reclaim_state;
>
>        /*
> @@ -2677,24 +2698,29 @@ int kswapd(void *p)
>                int new_classzone_idx;
>                int ret;
>
> -               new_order = pgdat->kswapd_max_order;
> -               new_classzone_idx = pgdat->classzone_idx;
> -               pgdat->kswapd_max_order = 0;
> -               pgdat->classzone_idx = MAX_NR_ZONES - 1;
> -               if (order < new_order || classzone_idx > new_classzone_idx) {
> -                       /*
> -                        * Don't sleep if someone wants a larger 'order'
> -                        * allocation or has tigher zone constraints
> -                        */
> -                       order = new_order;
> -                       classzone_idx = new_classzone_idx;
> -               } else {
> -                       kswapd_try_to_sleep(kswapd_p, order, classzone_idx);
> -                       order = pgdat->kswapd_max_order;
> -                       classzone_idx = pgdat->classzone_idx;
> +               if (is_node_kswapd(kswapd_p)) {
> +                       new_order = pgdat->kswapd_max_order;
> +                       new_classzone_idx = pgdat->classzone_idx;
>                        pgdat->kswapd_max_order = 0;
>                        pgdat->classzone_idx = MAX_NR_ZONES - 1;
> -               }
> +                       if (order < new_order ||
> +                                       classzone_idx > new_classzone_idx) {
> +                               /*
> +                                * Don't sleep if someone wants a larger 'order'
> +                                * allocation or has tigher zone constraints
> +                                */
> +                               order = new_order;
> +                               classzone_idx = new_classzone_idx;
> +                       } else {
> +                               kswapd_try_to_sleep(kswapd_p, order,
> +                                                   classzone_idx);
> +                               order = pgdat->kswapd_max_order;
> +                               classzone_idx = pgdat->classzone_idx;
> +                               pgdat->kswapd_max_order = 0;
> +                               pgdat->classzone_idx = MAX_NR_ZONES - 1;
> +                       }
> +               } else
> +                       kswapd_try_to_sleep(kswapd_p, order, classzone_idx);
>
>                ret = try_to_freeze();
>                if (kthread_should_stop())
> @@ -2705,8 +2731,13 @@ int kswapd(void *p)
>                 * after returning from the refrigerator
>                 */
>                if (!ret) {
> -                       trace_mm_vmscan_kswapd_wake(pgdat->node_id, order);
> -                       order = balance_pgdat(pgdat, order, &classzone_idx);
> +                       if (is_node_kswapd(kswapd_p)) {
> +                               trace_mm_vmscan_kswapd_wake(pgdat->node_id,
> +                                                               order);
> +                               order = balance_pgdat(pgdat, order,
> +                                                       &classzone_idx);
> +                       } else
> +                               balance_mem_cgroup_pgdat(mem, order);
>                }
>        }
>        return 0;
> @@ -2853,30 +2884,53 @@ static int __devinit cpu_callback(struct notifier_block *nfb,
>  * This kswapd start function will be called by init and node-hot-add.
>  * On node-hot-add, kswapd will moved to proper cpus if cpus are hot-added.
>  */
> -int kswapd_run(int nid)
> +int kswapd_run(int nid, struct mem_cgroup *mem)
>  {
> -       pg_data_t *pgdat = NODE_DATA(nid);
>        struct task_struct *kswapd_thr;
> +       pg_data_t *pgdat = NULL;
>        struct kswapd *kswapd_p;
> +       static char name[TASK_COMM_LEN];
> +       int memcg_id;
>        int ret = 0;
>
> -       if (pgdat->kswapd_wait)
> -               return 0;
> +       if (!mem) {
> +               pgdat = NODE_DATA(nid);
> +               if (pgdat->kswapd_wait)
> +                       return ret;
> +       }
>
>        kswapd_p = kzalloc(sizeof(struct kswapd), GFP_KERNEL);
>        if (!kswapd_p)
>                return -ENOMEM;
>
>        init_waitqueue_head(&kswapd_p->kswapd_wait);
> -       pgdat->kswapd_wait = &kswapd_p->kswapd_wait;
> -       kswapd_p->kswapd_pgdat = pgdat;
>
> -       kswapd_thr = kthread_run(kswapd, kswapd_p, "kswapd%d", nid);
> +       if (!mem) {
> +               pgdat->kswapd_wait = &kswapd_p->kswapd_wait;
> +               kswapd_p->kswapd_pgdat = pgdat;
> +               snprintf(name, TASK_COMM_LEN, "kswapd_%d", nid);
> +       } else {
> +               memcg_id = mem_cgroup_init_kswapd(mem, kswapd_p);
> +               if (!memcg_id) {
> +                       kfree(kswapd_p);
> +                       return ret;
> +               }
> +               snprintf(name, TASK_COMM_LEN, "memcg_%d", memcg_id);
> +       }
> +
> +       kswapd_thr = kthread_run(kswapd, kswapd_p, name);
>        if (IS_ERR(kswapd_thr)) {
>                /* failure at boot is fatal */
>                BUG_ON(system_state == SYSTEM_BOOTING);
> -               printk("Failed to start kswapd on node %d\n",nid);
> -               pgdat->kswapd_wait = NULL;
> +               if (!mem) {
> +                       printk(KERN_ERR "Failed to start kswapd on node %d\n",
> +                                                               nid);
> +                       pgdat->kswapd_wait = NULL;
> +               } else {
> +                       printk(KERN_ERR "Failed to start kswapd on memcg %d\n",
> +                                                               memcg_id);
> +                       mem_cgroup_clear_kswapd(mem);
> +               }
>                kfree(kswapd_p);
>                ret = -1;
>        } else
> @@ -2887,16 +2941,18 @@ int kswapd_run(int nid)
>  /*
>  * Called by memory hotplug when all memory in a node is offlined.
>  */
> -void kswapd_stop(int nid)
> +void kswapd_stop(int nid, struct mem_cgroup *mem)
>  {
>        struct task_struct *kswapd_thr = NULL;
>        struct kswapd *kswapd_p = NULL;
>        wait_queue_head_t *wait;
>
> -       pg_data_t *pgdat = NODE_DATA(nid);
> -
>        spin_lock(&kswapds_spinlock);
> -       wait = pgdat->kswapd_wait;
> +       if (!mem)
> +               wait = NODE_DATA(nid)->kswapd_wait;
> +       else
> +               wait = mem_cgroup_kswapd_wait(mem);
> +
>        if (wait) {
>                kswapd_p = container_of(wait, struct kswapd, kswapd_wait);
>                kswapd_thr = kswapd_p->kswapd_task;
> @@ -2916,7 +2972,7 @@ static int __init kswapd_init(void)
>
>        swap_setup();
>        for_each_node_state(nid, N_HIGH_MEMORY)
> -               kswapd_run(nid);
> +               kswapd_run(nid, NULL);
>        hotcpu_notifier(cpu_callback, 0);
>        return 0;
>  }
> --
> 1.7.3.1
>
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
> Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
>

  reply	other threads:[~2011-04-14  3:58 UTC|newest]

Thread overview: 27+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-04-13  7:03 [PATCH V3 0/7] memcg: per cgroup background reclaim Ying Han
2011-04-13  7:03 ` [PATCH V3 1/7] Add kswapd descriptor Ying Han
2011-04-13  7:03 ` [PATCH V3 2/7] Add per memcg reclaim watermarks Ying Han
2011-04-13  8:25   ` KAMEZAWA Hiroyuki
2011-04-13 18:40     ` Ying Han
2011-04-14  0:27       ` KAMEZAWA Hiroyuki
2011-04-14  8:24   ` Zhu Yanhai
2011-04-14 17:43     ` Ying Han
2011-04-13  7:03 ` [PATCH V3 3/7] New APIs to adjust per-memcg wmarks Ying Han
2011-04-13  8:30   ` KAMEZAWA Hiroyuki
2011-04-13 18:46     ` Ying Han
2011-04-13  7:03 ` [PATCH V3 4/7] Infrastructure to support per-memcg reclaim Ying Han
2011-04-14  3:57   ` Zhu Yanhai [this message]
2011-04-14  6:32     ` Ying Han
2011-04-13  7:03 ` [PATCH V3 5/7] Per-memcg background reclaim Ying Han
2011-04-13  8:58   ` KAMEZAWA Hiroyuki
2011-04-13 22:45     ` Ying Han
2011-04-13  7:03 ` [PATCH V3 6/7] Enable per-memcg " Ying Han
2011-04-13  9:05   ` KAMEZAWA Hiroyuki
2011-04-13 21:20     ` Ying Han
2011-04-14  0:30       ` KAMEZAWA Hiroyuki
2011-04-13  7:03 ` [PATCH V3 7/7] Add some per-memcg stats Ying Han
2011-04-13  7:47 ` [PATCH V3 0/7] memcg: per cgroup background reclaim KAMEZAWA Hiroyuki
2011-04-13 17:53   ` Ying Han
2011-04-14  0:14     ` KAMEZAWA Hiroyuki
2011-04-14 17:38       ` Ying Han
2011-04-14 21:59         ` Ying Han

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='BANLkTi=Qdv3RZuhN++OX2-S8OZqrL7=KBg@mail.gmail.com' \
    --to=zhu.yanhai@gmail.com \
    --cc=akpm@linux-foundation.org \
    --cc=balbir@linux.vnet.ibm.com \
    --cc=cl@linux.com \
    --cc=dave@linux.vnet.ibm.com \
    --cc=hannes@cmpxchg.org \
    --cc=hughd@google.com \
    --cc=kamezawa.hiroyu@jp.fujitsu.com \
    --cc=kosaki.motohiro@jp.fujitsu.com \
    --cc=linux-mm@kvack.org \
    --cc=lizf@cn.fujitsu.com \
    --cc=mel@csn.ul.ie \
    --cc=mhocko@suse.cz \
    --cc=nishimura@mxp.nes.nec.co.jp \
    --cc=riel@redhat.com \
    --cc=tj@kernel.org \
    --cc=xemul@openvz.org \
    --cc=yinghan@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).