mm/thp: move lru_add_page_tail() to huge_memory.c
authorAlex Shi <alex.shi@linux.alibaba.com>
Tue, 15 Dec 2020 20:33:20 +0000 (12:33 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 15 Dec 2020 22:48:03 +0000 (14:48 -0800)
Patch series "per memcg lru lock", v21.

This patchset includes 3 parts:

 1) some code cleanup and minimum optimization as preparation

 2) use TestCleanPageLRU as page isolation's precondition

 3) replace per node lru_lock with per memcg per node lru_lock

Current lru_lock is one for each of node, pgdat->lru_lock, that guard
for lru lists, but now we had moved the lru lists into memcg for long
time.  Still using per node lru_lock is clearly unscalable, pages on
each of memcgs have to compete each others for a whole lru_lock.  This
patchset try to use per lruvec/memcg lru_lock to repleace per node lru
lock to guard lru lists, make it scalable for memcgs and get performance
gain.

Currently lru_lock still guards both lru list and page's lru bit, that's
ok.  but if we want to use specific lruvec lock on the page, we need to
pin down the page's lruvec/memcg during locking.  Just taking lruvec
lock first may be undermined by the page's memcg charge/migration.  To
fix this problem, we could take out the page's lru bit clear and use it
as pin down action to block the memcg changes.  That's the reason for
new atomic func TestClearPageLRU.  So now isolating a page need both
actions: TestClearPageLRU and hold the lru_lock.

The typical usage of this is isolate_migratepages_block() in
compaction.c we have to take lru bit before lru lock, that serialized
the page isolation in memcg page charge/migration which will change
page's lruvec and new lru_lock in it.

The above solution suggested by Johannes Weiner, and based on his new
memcg charge path, then have this patchset.  (Hugh Dickins tested and
contributed much code from compaction fix to general code polish, thanks
a lot!).

Daniel Jordan's testing show 62% improvement on modified readtwice case
on his 2P * 10 core * 2 HT broadwell box on v18, which has no much
different with this v20.

 https://lore.kernel.org/lkml/20200915165807.kpp7uhiw7l3loofu@ca-dmjordan1.us.oracle.com/

Thanks to Hugh Dickins and Konstantin Khlebnikov, they both brought this
idea 8 years ago, and others who gave comments as well: Daniel Jordan,
Mel Gorman, Shakeel Butt, Matthew Wilcox, Alexander Duyck etc.

Thanks for Testing support from Intel 0day and Rong Chen, Fengguang Wu,
and Yun Wang.  Hugh Dickins also shared his kbuild-swap case.

This patch (of 19):

lru_add_page_tail() is only used in huge_memory.c, defining it in other
file with a CONFIG_TRANSPARENT_HUGEPAGE macro restrict just looks weird.

Let's move it THP. And make it static as Hugh Dickins suggested.

Link: https://lkml.kernel.org/r/1604566549-62481-1-git-send-email-alex.shi@linux.alibaba.com
Link: https://lkml.kernel.org/r/1604566549-62481-2-git-send-email-alex.shi@linux.alibaba.com
Signed-off-by: Alex Shi <alex.shi@linux.alibaba.com>
Reviewed-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Tejun Heo <tj@kernel.org>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Wei Yang <richard.weiyang@gmail.com>
Cc: Alexander Duyck <alexander.duyck@gmail.com>
Cc: "Chen, Rong A" <rong.a.chen@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: "Huang, Ying" <ying.huang@intel.com>
Cc: Jann Horn <jannh@google.com>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mika Penttilä <mika.penttila@nextfour.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/linux/swap.h
mm/huge_memory.c
mm/swap.c

index 667935c0dbd4c509261784b822dd3d7347f78d87..5e1e967c225f432b108076fac4f4b6d6f78d3721 100644 (file)
@@ -338,8 +338,6 @@ extern void lru_note_cost(struct lruvec *lruvec, bool file,
                          unsigned int nr_pages);
 extern void lru_note_cost_page(struct page *);
 extern void lru_cache_add(struct page *);
-extern void lru_add_page_tail(struct page *page, struct page *page_tail,
-                        struct lruvec *lruvec, struct list_head *head);
 extern void mark_page_accessed(struct page *);
 extern void lru_add_drain(void);
 extern void lru_add_drain_cpu(int cpu);
index 57d08156acb1d5d0674a4cb1a89afc89637f0f88..85b50baa7c7e57b617de56013b5328c888068f04 100644 (file)
@@ -2359,6 +2359,36 @@ static void remap_page(struct page *page, unsigned int nr)
        }
 }
 
+static void lru_add_page_tail(struct page *page, struct page *page_tail,
+               struct lruvec *lruvec, struct list_head *list)
+{
+       VM_BUG_ON_PAGE(!PageHead(page), page);
+       VM_BUG_ON_PAGE(PageCompound(page_tail), page);
+       VM_BUG_ON_PAGE(PageLRU(page_tail), page);
+       lockdep_assert_held(&lruvec_pgdat(lruvec)->lru_lock);
+
+       if (!list)
+               SetPageLRU(page_tail);
+
+       if (likely(PageLRU(page)))
+               list_add_tail(&page_tail->lru, &page->lru);
+       else if (list) {
+               /* page reclaim is reclaiming a huge page */
+               get_page(page_tail);
+               list_add_tail(&page_tail->lru, list);
+       } else {
+               /*
+                * Head page has not yet been counted, as an hpage,
+                * so we must account for each subpage individually.
+                *
+                * Put page_tail on the list at the correct position
+                * so they all end up in order.
+                */
+               add_page_to_lru_list_tail(page_tail, lruvec,
+                                         page_lru(page_tail));
+       }
+}
+
 static void __split_huge_page_tail(struct page *head, int tail,
                struct lruvec *lruvec, struct list_head *list)
 {
index 16a525296960d07ae735ac8f6a4ef3eed1db85b0..6e3121dd6e07c075fc2e7e959b49bda8f366de61 100644 (file)
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -977,39 +977,6 @@ void __pagevec_release(struct pagevec *pvec)
 }
 EXPORT_SYMBOL(__pagevec_release);
 
-#ifdef CONFIG_TRANSPARENT_HUGEPAGE
-/* used by __split_huge_page_refcount() */
-void lru_add_page_tail(struct page *page, struct page *page_tail,
-                      struct lruvec *lruvec, struct list_head *list)
-{
-       VM_BUG_ON_PAGE(!PageHead(page), page);
-       VM_BUG_ON_PAGE(PageCompound(page_tail), page);
-       VM_BUG_ON_PAGE(PageLRU(page_tail), page);
-       lockdep_assert_held(&lruvec_pgdat(lruvec)->lru_lock);
-
-       if (!list)
-               SetPageLRU(page_tail);
-
-       if (likely(PageLRU(page)))
-               list_add_tail(&page_tail->lru, &page->lru);
-       else if (list) {
-               /* page reclaim is reclaiming a huge page */
-               get_page(page_tail);
-               list_add_tail(&page_tail->lru, list);
-       } else {
-               /*
-                * Head page has not yet been counted, as an hpage,
-                * so we must account for each subpage individually.
-                *
-                * Put page_tail on the list at the correct position
-                * so they all end up in order.
-                */
-               add_page_to_lru_list_tail(page_tail, lruvec,
-                                         page_lru(page_tail));
-       }
-}
-#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
-
 static void __pagevec_lru_add_fn(struct page *page, struct lruvec *lruvec,
                                 void *arg)
 {