Re: [PATCH 4/9] kstaled: minimalistic implementation.

[Zhu Yanhai <zhu.yanhai@gmail.com>]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=UTF-8, Size: 14707 bytes --]

2011/9/28 Michel Lespinasse <walken@google.com>:
> Introduce minimal kstaled implementation. The scan rate is controlled by
> /sys/kernel/mm/kstaled/scan_seconds and per-cgroup statistics are output
> into /dev/cgroup/*/memory.idle_page_stats.
>
>
> Signed-off-by: Michel Lespinasse <walken@google.com>
> ---
>  mm/memcontrol.c |  297 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  1 files changed, 297 insertions(+), 0 deletions(-)
>
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index e013b8e..e55056f 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -49,6 +49,8 @@
>  #include <linux/page_cgroup.h>
>  #include <linux/cpu.h>
>  #include <linux/oom.h>
> +#include <linux/kthread.h>
> +#include <linux/rmap.h>
>  #include "internal.h"
>
>  #include <asm/uaccess.h>
> @@ -283,6 +285,16 @@ struct mem_cgroup {
>         */
>        struct mem_cgroup_stat_cpu nocpu_base;
>        spinlock_t pcp_counter_lock;
> +
> +#ifdef CONFIG_KSTALED
> +       seqcount_t idle_page_stats_lock;
> +       struct idle_page_stats {
> +               unsigned long idle_clean;
> +               unsigned long idle_dirty_file;
> +               unsigned long idle_dirty_swap;
> +       } idle_page_stats, idle_scan_stats;
> +       unsigned long idle_page_scans;
> +#endif
>  };
>
>  /* Stuffs for move charges at task migration. */
> @@ -4668,6 +4680,30 @@ static int mem_control_numa_stat_open(struct inode *unused, struct file *file)
>  }
>  #endif /* CONFIG_NUMA */
>
> +#ifdef CONFIG_KSTALED
> +static int mem_cgroup_idle_page_stats_read(struct cgroup *cgrp,
> +       struct cftype *cft,  struct cgroup_map_cb *cb)
> +{
> +       struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
> +       unsigned int seqcount;
> +       struct idle_page_stats stats;
> +       unsigned long scans;
> +
> +       do {
> +               seqcount = read_seqcount_begin(&memcg->idle_page_stats_lock);
> +               stats = memcg->idle_page_stats;
> +               scans = memcg->idle_page_scans;
> +       } while (read_seqcount_retry(&memcg->idle_page_stats_lock, seqcount));
> +
> +       cb->fill(cb, "idle_clean", stats.idle_clean * PAGE_SIZE);
> +       cb->fill(cb, "idle_dirty_file", stats.idle_dirty_file * PAGE_SIZE);
> +       cb->fill(cb, "idle_dirty_swap", stats.idle_dirty_swap * PAGE_SIZE);
> +       cb->fill(cb, "scans", scans);
> +
> +       return 0;
> +}
> +#endif /* CONFIG_KSTALED */
> +
>  static struct cftype mem_cgroup_files[] = {
>        {
>                .name = "usage_in_bytes",
> @@ -4738,6 +4774,12 @@ static struct cftype mem_cgroup_files[] = {
>                .mode = S_IRUGO,
>        },
>  #endif
> +#ifdef CONFIG_KSTALED
> +       {
> +               .name = "idle_page_stats",
> +               .read_map = mem_cgroup_idle_page_stats_read,
> +       },
> +#endif
>  };
>
>  #ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
> @@ -5001,6 +5043,9 @@ mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont)
>        atomic_set(&mem->refcnt, 1);
>        mem->move_charge_at_immigrate = 0;
>        mutex_init(&mem->thresholds_lock);
> +#ifdef CONFIG_KSTALED
> +       seqcount_init(&mem->idle_page_stats_lock);
> +#endif
>        return &mem->css;
>  free_out:
>        __mem_cgroup_free(mem);
> @@ -5568,3 +5613,255 @@ static int __init enable_swap_account(char *s)
>  __setup("swapaccount=", enable_swap_account);
>
>  #endif
> +
> +#ifdef CONFIG_KSTALED
> +
> +static unsigned int kstaled_scan_seconds;
> +static DECLARE_WAIT_QUEUE_HEAD(kstaled_wait);
> +
> +static unsigned kstaled_scan_page(struct page *page)
> +{
> +       bool is_locked = false;
> +       bool is_file;
> +       struct page_referenced_info info;
> +       struct page_cgroup *pc;
> +       struct idle_page_stats *stats;
> +       unsigned nr_pages;
> +
> +       /*
> +        * Before taking the page reference, check if the page is
> +        * a user page which is not obviously unreclaimable
> +        * (we will do more complete checks later).
> +        */
> +       if (!PageLRU(page) ||
> +           (!PageCompound(page) &&
> +            (PageMlocked(page) ||
> +             (page->mapping == NULL && !PageSwapCache(page)))))
> +               return 1;
> +
> +       if (!get_page_unless_zero(page))
> +               return 1;
> +
> +       /* Recheck now that we have the page reference. */
> +       if (unlikely(!PageLRU(page)))
> +               goto out;
> +       nr_pages = 1 << compound_trans_order(page);
> +       if (PageMlocked(page))
> +               goto out;
> +
> +       /*
> +        * Anon and SwapCache pages can be identified without locking.
> +        * For all other cases, we need the page locked in order to
> +        * dereference page->mapping.
> +        */
> +       if (PageAnon(page) || PageSwapCache(page))
> +               is_file = false;
> +       else if (!trylock_page(page)) {
> +               /*
> +                * We need to lock the page to dereference the mapping.
> +                * But don't risk sleeping by calling lock_page().
> +                * We don't want to stall kstaled, so we conservatively
> +                * count locked pages as unreclaimable.
> +                */
> +               goto out;
> +       } else {
> +               struct address_space *mapping = page->mapping;
> +
> +               is_locked = true;
> +
> +               /*
> +                * The page is still anon - it has been continuously referenced
> +                * since the prior check.
> +                */
> +               VM_BUG_ON(PageAnon(page) || mapping != page_rmapping(page));
> +
> +               /*
> +                * Check the mapping under protection of the page lock.
> +                * 1. If the page is not swap cache and has no mapping,
> +                *    shrink_page_list can't do anything with it.
> +                * 2. If the mapping is unevictable (as in SHM_LOCK segments),
> +                *    shrink_page_list can't do anything with it.
> +                * 3. If the page is swap cache or the mapping is swap backed
> +                *    (as in shmem), consider it a swappable page.
> +                * 4. If the backing dev has indicated that it does not want
> +                *    its pages sync'd to disk (as in ramfs), take this as
> +                *    a hint that its pages are not reclaimable.
> +                * 5. Otherwise, consider this as a file page reclaimable
> +                *    through standard pageout.
> +                */
> +               if (!mapping && !PageSwapCache(page))
> +                       goto out;
> +               else if (mapping_unevictable(mapping))
> +                       goto out;
> +               else if (PageSwapCache(page) ||
> +                        mapping_cap_swap_backed(mapping))
> +                       is_file = false;
> +               else if (!mapping_cap_writeback_dirty(mapping))
> +                       goto out;
> +               else
> +                       is_file = true;
> +       }
> +
> +       /* Find out if the page is idle. Also test for pending mlock. */
> +       page_referenced_kstaled(page, is_locked, &info);
> +       if ((info.pr_flags & PR_REFERENCED) || (info.vm_flags & VM_LOCKED))
> +               goto out;
> +
> +       /* Locate kstaled stats for the page's cgroup. */
> +       pc = lookup_page_cgroup(page);
> +       if (!pc)
> +               goto out;
> +       lock_page_cgroup(pc);
> +       if (!PageCgroupUsed(pc))
> +               goto unlock_page_cgroup_out;
> +       stats = &pc->mem_cgroup->idle_scan_stats;
Is it safe to deference it like this? I think we need something like this:
struct mem_cgroup *memcg = pc->mem_cgroup;
if (!memcg || !css_tryget(&memcg->css))
   goto out;
And also css_put() in soewhere bmelow.
Or simply remove the lock_page_cgroup() above and use
try_get_mem_cgroup_from_page() directly.

--
Thanks,
Zhu Yanhai
> +
> +       /* Finally increment the correct statistic for this page. */
> +       if (!(info.pr_flags & PR_DIRTY) &&
> +           !PageDirty(page) && !PageWriteback(page))
> +               stats->idle_clean += nr_pages;
> +       else if (is_file)
> +               stats->idle_dirty_file += nr_pages;
> +       else
> +               stats->idle_dirty_swap += nr_pages;
> +
> + unlock_page_cgroup_out:
> +       unlock_page_cgroup(pc);
> +
> + out:
> +       if (is_locked)
> +               unlock_page(page);
> +       put_page(page);
> +
> +       return nr_pages;
> +}
> +
> +static void kstaled_scan_node(pg_data_t *pgdat)
> +{
> +       unsigned long flags;
> +       unsigned long pfn, end;
> +
> +       pgdat_resize_lock(pgdat, &flags);
> +
> +       pfn = pgdat->node_start_pfn;
> +       end = pfn + pgdat->node_spanned_pages;
> +
> +       while (pfn < end) {
> +               if (need_resched()) {
> +                       pgdat_resize_unlock(pgdat, &flags);
> +                       cond_resched();
> +                       pgdat_resize_lock(pgdat, &flags);
> +
> +#ifdef CONFIG_MEMORY_HOTPLUG
> +                       /* abort if the node got resized */
> +                       if (pfn < pgdat->node_start_pfn ||
> +                           end > (pgdat->node_start_pfn +
> +                                  pgdat->node_spanned_pages))
> +                               goto abort;
> +#endif
> +               }
> +
> +               pfn += pfn_valid(pfn) ?
> +                       kstaled_scan_page(pfn_to_page(pfn)) : 1;
> +       }
> +
> +abort:
> +       pgdat_resize_unlock(pgdat, &flags);
> +}
> +
> +static int kstaled(void *dummy)
> +{
> +       while (1) {
> +               int scan_seconds;
> +               int nid;
> +               struct mem_cgroup *memcg;
> +
> +               wait_event_interruptible(kstaled_wait,
> +                                (scan_seconds = kstaled_scan_seconds) > 0);
> +               /*
> +                * We use interruptible wait_event so as not to contribute
> +                * to the machine load average while we're sleeping.
> +                * However, we don't actually expect to receive a signal
> +                * since we run as a kernel thread, so the condition we were
> +                * waiting for should be true once we get here.
> +                */
> +               BUG_ON(scan_seconds <= 0);
> +
> +               for_each_mem_cgroup_all(memcg)
> +                       memset(&memcg->idle_scan_stats, 0,
> +                              sizeof(memcg->idle_scan_stats));
> +
> +               for_each_node_state(nid, N_HIGH_MEMORY)
> +                       kstaled_scan_node(NODE_DATA(nid));
> +
> +               for_each_mem_cgroup_all(memcg) {
> +                       write_seqcount_begin(&memcg->idle_page_stats_lock);
> +                       memcg->idle_page_stats = memcg->idle_scan_stats;
> +                       memcg->idle_page_scans++;
> +                       write_seqcount_end(&memcg->idle_page_stats_lock);
> +               }
> +
> +               schedule_timeout_interruptible(scan_seconds * HZ);
> +       }
> +
> +       BUG();
> +       return 0;       /* NOT REACHED */
> +}
> +
> +static ssize_t kstaled_scan_seconds_show(struct kobject *kobj,
> +                                        struct kobj_attribute *attr,
> +                                        char *buf)
> +{
> +       return sprintf(buf, "%u\n", kstaled_scan_seconds);
> +}
> +
> +static ssize_t kstaled_scan_seconds_store(struct kobject *kobj,
> +                                         struct kobj_attribute *attr,
> +                                         const char *buf, size_t count)
> +{
> +       int err;
> +       unsigned long input;
> +
> +       err = kstrtoul(buf, 10, &input);
> +       if (err)
> +               return -EINVAL;
> +       kstaled_scan_seconds = input;
> +       wake_up_interruptible(&kstaled_wait);
> +       return count;
> +}
> +
> +static struct kobj_attribute kstaled_scan_seconds_attr = __ATTR(
> +       scan_seconds, 0644,
> +       kstaled_scan_seconds_show, kstaled_scan_seconds_store);
> +
> +static struct attribute *kstaled_attrs[] = {
> +       &kstaled_scan_seconds_attr.attr,
> +       NULL
> +};
> +static struct attribute_group kstaled_attr_group = {
> +       .name = "kstaled",
> +       .attrs = kstaled_attrs,
> +};
> +
> +static int __init kstaled_init(void)
> +{
> +       int error;
> +       struct task_struct *thread;
> +
> +       error = sysfs_create_group(mm_kobj, &kstaled_attr_group);
> +       if (error) {
> +               pr_err("Failed to create kstaled sysfs node\n");
> +               return error;
> +       }
> +
> +       thread = kthread_run(kstaled, NULL, "kstaled");
> +       if (IS_ERR(thread)) {
> +               pr_err("Failed to start kstaled\n");
> +               return PTR_ERR(thread);
> +       }
> +
> +       return 0;
> +}
> +module_init(kstaled_init);
> +
> +#endif /* CONFIG_KSTALED */
> --
> 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>
N‹§²æìr¸›zǧu©ž²Æ {\b­†éì¹»\x1c®&Þ–)îÆi¢žØ^n‡r¶‰šŽŠÝ¢j$½§$¢¸\x05¢¹¨­è§~Š'.)îÄÃ,yèm¶ŸÿÃ\f%Š{±šj+ƒñb‚^[nö¢®×¥yÊ&Š{^®w­r\x16«ë"œ&§iÖ¬Š	á¶Ú\x7fþËh¦Ø^™ë^­Æ¿\x0e‰ízf¢•¨ky