Skip to contentSkip to conversation

Introduce MADV_COLD and MADV_PAGEOUT

PATCH Superseded Minchan Kim 4 participants 27 messages Jun 27, 2019 – Jul 11, 2019

Subsystems (16)

v3 of 7 Superseded Latest is v7 →

lore
Share & tools
lore
#0 lore raw

v3 cover Introduce MADV_COLD and MADV_PAGEOUT

This patch is part of previous series:
https://lore.kernel.org/lkml/[email protected]/T/#u
Originally, it was created for external madvise hinting feature.

https://lkml.org/lkml/2019/5/31/463
Michal wanted to separte the discussion from external hinting interface so this patchset includes only first part of my entire patchset

- introduce MADV_COLD and MADV_PAGEOUT hint to madvise.

However, I keep entire description for others for easier understanding why this kinds of hint was born.

Thanks.

This patchset is against on next-20190530.

Below is description of previous entire patchset.
================= &< =====================

- Background

The Android terminology used for forking a new process and starting an app from scratch is a cold start, while resuming an existing app is a hot start. While we continually try to improve the performance of cold starts, hot starts will always be significantly less power hungry as well as faster so we are trying to make hot start more likely than cold start.

To increase hot start, Android userspace manages the order that apps should be killed in a process called ActivityManagerService. ActivityManagerService tracks every Android app or service that the user could be interacting with at any time and translates that into a ranked list for lmkd(low memory killer daemon). They are likely to be killed by lmkd if the system has to reclaim memory. In that sense they are similar to entries in any other cache. Those apps are kept alive for opportunistic performance improvements but those performance improvements will vary based on the memory requirements of individual workloads.

- Problem

Naturally, cached apps were dominant consumers of memory on the system. However, they were not significant consumers of swap even though they are good candidate for swap. Under investigation, swapping out only begins once the low zone watermark is hit and kswapd wakes up, but the overall allocation rate in the system might trip lmkd thresholds and cause a cached process to be killed(we measured performance swapping out vs. zapping the memory by killing a process. Unsurprisingly, zapping is 10x times faster even though we use zram which is much faster than real storage) so kill from lmkd will often satisfy the high zone watermark, resulting in very few pages actually being moved to swap.

- Approach

The approach we chose was to use a new interface to allow userspace to proactively reclaim entire processes by leveraging platform information. This allowed us to bypass the inaccuracy of the kernel’s LRUs for pages that are known to be cold from userspace and to avoid races with lmkd by reclaiming apps as soon as they entered the cached state. Additionally, it could provide many chances for platform to use much information to optimize memory efficiency.

To achieve the goal, the patchset introduce two new options for madvise. One is MADV_COLD which will deactivate activated pages and the other is MADV_PAGEOUT which will reclaim private pages instantly. These new options complement MADV_DONTNEED and MADV_FREE by adding non-destructive ways to gain some free memory space. MADV_PAGEOUT is similar to MADV_DONTNEED in a way that it hints the kernel that memory region is not currently needed and should be reclaimed immediately; MADV_COLD is similar to MADV_FREE in a way that it hints the kernel that memory region is not currently needed and should be reclaimed when memory pressure rises.

Minchan Kim (5):
mm: introduce MADV_COLD
mm: change PAGEREF_RECLAIM_CLEAN with PAGE_REFRECLAIM
mm: account nr_isolated_xxx in [isolate|putback]_lru_page
mm: introduce MADV_PAGEOUT
mm: factor out pmd young/dirty bit handling and THP split

 include/linux/huge_mm.h                |   3 -
 include/linux/swap.h                   |   2 +
 include/uapi/asm-generic/mman-common.h |   2 +
 mm/compaction.c                        |   2 -
 mm/gup.c                               |   7 +-
 mm/huge_memory.c                       |  74 -----
 mm/internal.h                          |   2 +-
 mm/khugepaged.c                        |   3 -
 mm/madvise.c                           | 438 ++++++++++++++++++++++++-
 mm/memory-failure.c                    |   3 -
 mm/memory_hotplug.c                    |   4 -
 mm/mempolicy.c                         |   6 +-
 mm/migrate.c                           |  37 +--
 mm/oom_kill.c                          |   2 +-
 mm/swap.c                              |  42 +++
 mm/vmscan.c                            |  86 ++++-
 16 files changed, 566 insertions(+), 147 deletions(-)
Signature · 1 line
2.22.0.410.gd8fdbe21b5-goog
#1 lore raw

PATCH v3 1/5 mm: introduce MADV_COLD

When a process expects no accesses to a certain memory range, it could give a hint to kernel that the pages can be reclaimed when memory pressure happens but data should be preserved for future use. This could reduce workingset eviction so it ends up increasing performance.

This patch introduces the new MADV_COLD hint to madvise(2) syscall. MADV_COLD can be used by a process to mark a memory range as not expected to be used in the near future. The hint can help kernel in deciding which pages to evict early during memory pressure.

It works for every LRU pages like MADV_[DONTNEED|FREE]. IOW, It moves

	active file page -> inactive file LRU
	active anon page -> inacdtive anon LRU

Unlike MADV_FREE, it doesn't move active anonymous pages to inactive file LRU's head because MADV_COLD is a little bit different symantic. MADV_FREE means it's okay to discard when the memory pressure because the content of the page is *garbage* so freeing such pages is almost zero overhead since we don't need to swap out and access afterward causes just minor fault. Thus, it would make sense to put those freeable pages in inactive file LRU to compete other used-once pages. It makes sense for implmentaion point of view, too because it's not swapbacked memory any longer until it would be re-dirtied. Even, it could give a bonus to make them be reclaimed on swapless system. However, MADV_COLD doesn't mean garbage so reclaiming them requires swap-out/in in the end so it's bigger cost. Since we have designed VM LRU aging based on cost-model, anonymous cold pages would be better to position inactive anon's LRU list, not file LRU. Furthermore, it would help to avoid unnecessary scanning if system doesn't have a swap device. Let's start simpler way without adding complexity at this moment. However, keep in mind, too that it's a caveat that workloads with a lot of pages cache are likely to ignore MADV_COLD on anonymous memory because we rarely age anonymous LRU lists.

* man-page material

MADV_COLD (since Linux x.x)

Do not expect access in the near future so under memory pressure, pages in the specified regions could be reclaimed more aggressively compared to other pages in the system. The difference with MADV_DONTNEED is it doesn't change the semantics of memory access in the specified regions. Thus, it will keep the up-to-date contents of the region.

MADV_COLD cannot be applied to locked pages, Huge TLB pages, or VM_PFNMAP pages.

* v2
* add up the warn with lots of page cache workload - mhocko
* add man page stuff - dave

* v1
* remove page_mapcount filter - hannes, mhocko
* remove idle page handling - joelaf

* RFCv2
* add more description - mhocko

* RFCv1
* renaming from MADV_COOL to MADV_COLD - hannes

* internal review
* use clear_page_youn in deactivate_page - joelaf
* Revise the description - surenb
* Renaming from MADV_WARM to MADV_COOL - surenb

---
 include/linux/swap.h                   |   1 +
 include/uapi/asm-generic/mman-common.h |   1 +
 mm/internal.h                          |   2 +-
 mm/madvise.c                           | 180 ++++++++++++++++++++++++-
 mm/oom_kill.c                          |   2 +-
 mm/swap.c                              |  42 ++++++
 6 files changed, 224 insertions(+), 4 deletions(-)
include/linux/swap.h +1 -0
index de2c67a33b7e..0ce997edb8bb 100644
@@ -340,6 +340,7 @@ extern void lru_add_drain_cpu(int cpu);
 extern void lru_add_drain_all(void);
 extern void rotate_reclaimable_page(struct page *page);
 extern void deactivate_file_page(struct page *page);
+extern void deactivate_page(struct page *page);
 extern void mark_page_lazyfree(struct page *page);
 extern void swap_setup(void);
 
include/uapi/asm-generic/mman-common.h +1 -0
index ef4623f03156..d7b4231eea63 100644
@@ -47,6 +47,7 @@
 #define MADV_SEQUENTIAL	2		/* expect sequential page references */
 #define MADV_WILLNEED	3		/* will need these pages */
 #define MADV_DONTNEED	4		/* don't need these pages */
+#define MADV_COLD	5		/* deactivatie these pages */
 
 /* common parameters: try to keep these consistent across architectures */
 #define MADV_FREE	8		/* free pages only if memory pressure */
mm/internal.h +1 -1
index f53a14d67538..c61b215ff265 100644
@@ -39,7 +39,7 @@ vm_fault_t do_swap_page(struct vm_fault *vmf);
 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
 		unsigned long floor, unsigned long ceiling);
 
-static inline bool can_madv_dontneed_vma(struct vm_area_struct *vma)
+static inline bool can_madv_lru_vma(struct vm_area_struct *vma)
 {
 	return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP));
 }
mm/madvise.c +178 -2

216 lines not on this page · raw message

mm/oom_kill.c +1 -1
index 6de5c354d6ca..2140a6f8db63 100644
@@ -523,7 +523,7 @@ bool __oom_reap_task_mm(struct mm_struct *mm)
 	set_bit(MMF_UNSTABLE, &mm->flags);
 
 	for (vma = mm->mmap ; vma; vma = vma->vm_next) {
-		if (!can_madv_dontneed_vma(vma))
+		if (!can_madv_lru_vma(vma))
 			continue;
 
 		/*
mm/swap.c +42 -0

72 lines not on this page · raw message

Signature · 1 line
2.22.0.410.gd8fdbe21b5-goog
#10 lore raw

PATCH v3 2/5 mm: change PAGEREF_RECLAIM_CLEAN with PAGE_REFRECLAIM

In reply to Minchan Kim

The local variable references in shrink_page_list is PAGEREF_RECLAIM_CLEAN as default. It is for preventing to reclaim dirty pages when CMA try to migrate pages. Strictly speaking, we don't need it because CMA didn't allow to write out by .may_writepage = 0 in reclaim_clean_pages_from_list.

Moreover, it has a problem to prevent anonymous pages's swap out even though force_reclaim = true in shrink_page_list on upcoming patch. So this patch makes references's default value to PAGEREF_RECLAIM and rename force_reclaim with ignore_references to make it more clear.

This is a preparatory work for next patch.

* RFCv1
* use ignore_referecnes as parameter name - hannes

---
 mm/vmscan.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
mm/vmscan.c +3 -3
index 9e3292ee5c7c..49e9ee4d771d 100644
@@ -1117,7 +1117,7 @@ static unsigned long shrink_page_list(struct list_head *page_list,
 				      struct scan_control *sc,
 				      enum ttu_flags ttu_flags,
 				      struct reclaim_stat *stat,
-				      bool force_reclaim)
+				      bool ignore_references)
 {
 	LIST_HEAD(ret_pages);
 	LIST_HEAD(free_pages);
@@ -1131,7 +1131,7 @@ static unsigned long shrink_page_list(struct list_head *page_list,
 		struct address_space *mapping;
 		struct page *page;
 		int may_enter_fs;
-		enum page_references references = PAGEREF_RECLAIM_CLEAN;
+		enum page_references references = PAGEREF_RECLAIM;
 		bool dirty, writeback;
 		unsigned int nr_pages;
 
@@ -1262,7 +1262,7 @@ static unsigned long shrink_page_list(struct list_head *page_list,
 			}
 		}
 
-		if (!force_reclaim)
+		if (!ignore_references)
 			references = page_check_references(page, sc);
 
 		switch (references) {
Signature · 1 line
2.22.0.410.gd8fdbe21b5-goog
#11 lore raw

PATCH v3 3/5 mm: account nr_isolated_xxx in [isolate|putback]_lru_page

In reply to Minchan Kim

The isolate counting is pecpu counter so it would be not huge gain to work them by batch. Rather than complicating to make them batch, let's make it more stright-foward via adding the counting logic into [isolate|putback]_lru_page API.

* v1
* fix accounting bug - Hillf

---
 mm/compaction.c     |  2 --
 mm/gup.c            |  7 +------
 mm/khugepaged.c     |  3 ---
 mm/memory-failure.c |  3 ---
 mm/memory_hotplug.c |  4 ----
 mm/mempolicy.c      |  6 +-----
 mm/migrate.c        | 37 ++++++++-----------------------------
 mm/vmscan.c         | 22 ++++++++++++++++------
 8 files changed, 26 insertions(+), 58 deletions(-)
mm/compaction.c +0 -2
index 9e1b9acb116b..c6591682deda 100644
@@ -982,8 +982,6 @@ isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn,
 
 		/* Successfully isolated */
 		del_page_from_lru_list(page, lruvec, page_lru(page));
-		inc_node_page_state(page,
-				NR_ISOLATED_ANON + page_is_file_cache(page));
 
 isolate_success:
 		list_add(&page->lru, &cc->migratepages);
mm/gup.c +1 -6
index 7dde2e3a1963..aec3a2b7e61b 100644
@@ -1473,13 +1473,8 @@ static long check_and_migrate_cma_pages(struct task_struct *tsk,
 					drain_allow = false;
 				}
 
-				if (!isolate_lru_page(head)) {
+				if (!isolate_lru_page(head))
 					list_add_tail(&head->lru, &cma_page_list);
-					mod_node_page_state(page_pgdat(head),
-							    NR_ISOLATED_ANON +
-							    page_is_file_cache(head),
-							    hpage_nr_pages(head));
-				}
 			}
 		}
 	}
mm/khugepaged.c +0 -3
index 0f7419938008..7da34e198ec5 100644
@@ -503,7 +503,6 @@ void __khugepaged_exit(struct mm_struct *mm)
 
 static void release_pte_page(struct page *page)
 {
-	dec_node_page_state(page, NR_ISOLATED_ANON + page_is_file_cache(page));
 	unlock_page(page);
 	putback_lru_page(page);
 }
@@ -602,8 +601,6 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 			result = SCAN_DEL_PAGE_LRU;
 			goto out;
 		}
-		inc_node_page_state(page,
-				NR_ISOLATED_ANON + page_is_file_cache(page));
 		VM_BUG_ON_PAGE(!PageLocked(page), page);
 		VM_BUG_ON_PAGE(PageLRU(page), page);
 
mm/memory-failure.c +0 -3
index 7e08cbf3ba49..3586e8226e4e 100644
@@ -1795,9 +1795,6 @@ static int __soft_offline_page(struct page *page, int flags)
 		 * so use !__PageMovable instead for LRU page's mapping
 		 * cannot have PAGE_MAPPING_MOVABLE.
 		 */
-		if (!__PageMovable(page))
-			inc_node_page_state(page, NR_ISOLATED_ANON +
-						page_is_file_cache(page));
 		list_add(&page->lru, &pagelist);
 		ret = migrate_pages(&pagelist, new_page, NULL, MPOL_MF_MOVE_ALL,
 					MIGRATE_SYNC, MR_MEMORY_FAILURE);
mm/memory_hotplug.c +0 -4
index dfab21dc33dc..68577c677b46 100644
@@ -1384,10 +1384,6 @@ do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
 			ret = isolate_movable_page(page, ISOLATE_UNEVICTABLE);
 		if (!ret) { /* Success */
 			list_add_tail(&page->lru, &source);
-			if (!__PageMovable(page))
-				inc_node_page_state(page, NR_ISOLATED_ANON +
-						    page_is_file_cache(page));
-
 		} else {
 			pr_warn("failed to isolate pfn %lx\n", pfn);
 			dump_page(page, "isolation failed");
mm/mempolicy.c +1 -5
index 64562809bf3b..03081f3404ca 100644
@@ -994,12 +994,8 @@ static int migrate_page_add(struct page *page, struct list_head *pagelist,
 	 * Avoid migrating a page that is shared with others.
 	 */
 	if ((flags & MPOL_MF_MOVE_ALL) || page_mapcount(head) == 1) {
-		if (!isolate_lru_page(head)) {
+		if (!isolate_lru_page(head))
 			list_add_tail(&head->lru, pagelist);
-			mod_node_page_state(page_pgdat(head),
-				NR_ISOLATED_ANON + page_is_file_cache(head),
-				hpage_nr_pages(head));
-		}
 	}
 
 	return 0;
mm/migrate.c +8 -29
index 572b4bc85d76..5583324c01e7 100644
@@ -190,8 +190,6 @@ void putback_movable_pages(struct list_head *l)
 			unlock_page(page);
 			put_page(page);
 		} else {
-			mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON +
-					page_is_file_cache(page), -hpage_nr_pages(page));
 			putback_lru_page(page);
 		}
 	}
@@ -1181,10 +1179,17 @@ static ICE_noinline int unmap_and_move(new_page_t get_new_page,
 		return -ENOMEM;
 
 	if (page_count(page) == 1) {
+		bool is_lru = !__PageMovable(page);
+
 		/* page was freed from under us. So we are done. */
 		ClearPageActive(page);
 		ClearPageUnevictable(page);
-		if (unlikely(__PageMovable(page))) {
+		if (likely(is_lru))
+			mod_node_page_state(page_pgdat(page),
+						NR_ISOLATED_ANON +
+						page_is_file_cache(page),
+						-hpage_nr_pages(page));
+		else {
 			lock_page(page);
 			if (!PageMovable(page))
 				__ClearPageIsolated(page);
@@ -1210,15 +1215,6 @@ static ICE_noinline int unmap_and_move(new_page_t get_new_page,
 		 * restored.
 		 */
 		list_del(&page->lru);
-
-		/*
-		 * Compaction can migrate also non-LRU pages which are
-		 * not accounted to NR_ISOLATED_*. They can be recognized
-		 * as __PageMovable
-		 */
-		if (likely(!__PageMovable(page)))
-			mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON +
-					page_is_file_cache(page), -hpage_nr_pages(page));
 	}
 
 	/*
@@ -1572,9 +1568,6 @@ static int add_page_for_migration(struct mm_struct *mm, unsigned long addr,
 
 		err = 0;
 		list_add_tail(&head->lru, pagelist);
-		mod_node_page_state(page_pgdat(head),
-			NR_ISOLATED_ANON + page_is_file_cache(head),
-			hpage_nr_pages(head));
 	}
 out_putpage:
 	/*
@@ -1890,8 +1883,6 @@ static struct page *alloc_misplaced_dst_page(struct page *page,
 
 static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
 {
-	int page_lru;
-
 	VM_BUG_ON_PAGE(compound_order(page) && !PageTransHuge(page), page);
 
 	/* Avoid migrating to a node that is nearly full */
@@ -1913,10 +1904,6 @@ static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
 		return 0;
 	}
 
-	page_lru = page_is_file_cache(page);
-	mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON + page_lru,
-				hpage_nr_pages(page));
-
 	/*
 	 * Isolating the page has taken another reference, so the
 	 * caller's reference can be safely dropped without the page
@@ -1971,8 +1958,6 @@ int migrate_misplaced_page(struct page *page, struct vm_area_struct *vma,
 	if (nr_remaining) {
 		if (!list_empty(&migratepages)) {
 			list_del(&page->lru);
-			dec_node_page_state(page, NR_ISOLATED_ANON +
-					page_is_file_cache(page));
 			putback_lru_page(page);
 		}
 		isolated = 0;
@@ -2002,7 +1987,6 @@ int migrate_misplaced_transhuge_page(struct mm_struct *mm,
 	pg_data_t *pgdat = NODE_DATA(node);
 	int isolated = 0;
 	struct page *new_page = NULL;
-	int page_lru = page_is_file_cache(page);
 	unsigned long start = address & HPAGE_PMD_MASK;
 
 	new_page = alloc_pages_node(node,
@@ -2048,8 +2032,6 @@ int migrate_misplaced_transhuge_page(struct mm_struct *mm,
 		/* Retake the callers reference and putback on LRU */
 		get_page(page);
 		putback_lru_page(page);
-		mod_node_page_state(page_pgdat(page),
-			 NR_ISOLATED_ANON + page_lru, -HPAGE_PMD_NR);
 
 		goto out_unlock;
 	}
@@ -2099,9 +2081,6 @@ int migrate_misplaced_transhuge_page(struct mm_struct *mm,
 	count_vm_events(PGMIGRATE_SUCCESS, HPAGE_PMD_NR);
 	count_vm_numa_events(NUMA_PAGE_MIGRATE, HPAGE_PMD_NR);
 
-	mod_node_page_state(page_pgdat(page),
-			NR_ISOLATED_ANON + page_lru,
-			-HPAGE_PMD_NR);
 	return isolated;
 
 out_fail:
mm/vmscan.c +16 -6
index 49e9ee4d771d..223ce5da08f0 100644
@@ -1014,6 +1014,9 @@ int remove_mapping(struct address_space *mapping, struct page *page)
 void putback_lru_page(struct page *page)
 {
 	lru_cache_add(page);
+	mod_node_page_state(page_pgdat(page),
+				NR_ISOLATED_ANON + page_is_file_cache(page),
+				-hpage_nr_pages(page));
 	put_page(page);		/* drop ref from isolate */
 }
 
@@ -1479,6 +1482,9 @@ static unsigned long shrink_page_list(struct list_head *page_list,
 		 */
 		nr_reclaimed += nr_pages;
 
+		mod_node_page_state(pgdat, NR_ISOLATED_ANON +
+						page_is_file_cache(page),
+						-nr_pages);
 		/*
 		 * Is there need to periodically free_page_list? It would
 		 * appear not as the counts should be low
@@ -1554,7 +1560,6 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 	ret = shrink_page_list(&clean_pages, zone->zone_pgdat, &sc,
 			TTU_IGNORE_ACCESS, &dummy_stat, true);
 	list_splice(&clean_pages, page_list);
-	mod_node_page_state(zone->zone_pgdat, NR_ISOLATED_FILE, -ret);
 	return ret;
 }
 
@@ -1630,6 +1635,9 @@ int __isolate_lru_page(struct page *page, isolate_mode_t mode)
 		 */
 		ClearPageLRU(page);
 		ret = 0;
+		__mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON +
+						page_is_file_cache(page),
+						hpage_nr_pages(page));
 	}
 
 	return ret;
@@ -1761,6 +1769,7 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
 	trace_mm_vmscan_lru_isolate(sc->reclaim_idx, sc->order, nr_to_scan,
 				    total_scan, skipped, nr_taken, mode, lru);
 	update_lru_sizes(lruvec, lru, nr_zone_taken);
+
 	return nr_taken;
 }
 
@@ -1809,6 +1818,9 @@ int isolate_lru_page(struct page *page)
 			ClearPageLRU(page);
 			del_page_from_lru_list(page, lruvec, lru);
 			ret = 0;
+			mod_node_page_state(pgdat, NR_ISOLATED_ANON +
+						page_is_file_cache(page),
+						hpage_nr_pages(page));
 		}
 		spin_unlock_irq(&pgdat->lru_lock);
 	}
@@ -1900,6 +1912,9 @@ static unsigned noinline_for_stack move_pages_to_lru(struct lruvec *lruvec,
 		update_lru_size(lruvec, lru, page_zonenum(page), nr_pages);
 		list_move(&page->lru, &lruvec->lists[lru]);
 
+		__mod_node_page_state(pgdat, NR_ISOLATED_ANON +
+						page_is_file_cache(page),
+						-hpage_nr_pages(page));
 		if (put_page_testzero(page)) {
 			__ClearPageLRU(page);
 			__ClearPageActive(page);
@@ -1977,7 +1992,6 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
 	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &page_list,
 				     &nr_scanned, sc, lru);
 
-	__mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, nr_taken);
 	reclaim_stat->recent_scanned[file] += nr_taken;
 
 	item = current_is_kswapd() ? PGSCAN_KSWAPD : PGSCAN_DIRECT;
@@ -2003,8 +2017,6 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
 
 	move_pages_to_lru(lruvec, &page_list);
 
-	__mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, -nr_taken);
-
 	spin_unlock_irq(&pgdat->lru_lock);
 
 	mem_cgroup_uncharge_list(&page_list);
@@ -2063,7 +2075,6 @@ static void shrink_active_list(unsigned long nr_to_scan,
 	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &l_hold,
 				     &nr_scanned, sc, lru);
 
-	__mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, nr_taken);
 	reclaim_stat->recent_scanned[file] += nr_taken;
 
 	__count_vm_events(PGREFILL, nr_scanned);
@@ -2132,7 +2143,6 @@ static void shrink_active_list(unsigned long nr_to_scan,
 	__count_vm_events(PGDEACTIVATE, nr_deactivate);
 	__count_memcg_events(lruvec_memcg(lruvec), PGDEACTIVATE, nr_deactivate);
 
-	__mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, -nr_taken);
 	spin_unlock_irq(&pgdat->lru_lock);
 
 	mem_cgroup_uncharge_list(&l_active);
Signature · 1 line
2.22.0.410.gd8fdbe21b5-goog
#13 lore raw

PATCH v3 4/5 mm: introduce MADV_PAGEOUT

In reply to Minchan Kim

When a process expects no accesses to a certain memory range for a long time, it could hint kernel that the pages can be reclaimed instantly but data should be preserved for future use. This could reduce workingset eviction so it ends up increasing performance.

This patch introduces the new MADV_PAGEOUT hint to madvise(2) syscall. MADV_PAGEOUT can be used by a process to mark a memory range as not expected to be used for a long time so that kernel reclaims *any LRU* pages instantly. The hint can help kernel in deciding which pages to evict proactively.

- man-page material

MADV_PAGEOUT (since Linux x.x)

Do not expect access in the near future so pages in the specified regions could be reclaimed instantly regardless of memory pressure. Thus, access in the range after successful operation could cause major page fault but never lose the up-to-date contents unlike MADV_DONTNEED. It works for only private anonymous mappings and non-anonymous mappings that belong to files that the calling process could successfully open for writing; otherwise, it could be used for sidechannel attack.

MADV_PAGEOUT cannot be applied to locked pages, Huge TLB pages, or VM_PFNMAP pages.

* v2
* add comment about SWAP_CLUSTER_MAX - mhocko
* add permission check to prevent sidechannel attack - mhocko
* add man page stuff - dave

* v1
* change pte to old and rely on the other's reference - hannes
* remove page_mapcount to check shared page - mhocko

* RFC v2
* make reclaim_pages simple via factoring out isolate logic - hannes

* RFCv1
* rename from MADV_COLD to MADV_PAGEOUT - hannes
* bail out if process is being killed - Hillf
* fix reclaim_pages bugs - Hillf

---
 include/linux/swap.h                   |   1 +
 include/uapi/asm-generic/mman-common.h |   1 +
 mm/madvise.c                           | 212 +++++++++++++++++++++++++
 mm/vmscan.c                            |  58 +++++++
 4 files changed, 272 insertions(+)
include/linux/swap.h +1 -0
index 0ce997edb8bb..063c0c1e112b 100644
@@ -365,6 +365,7 @@ extern int vm_swappiness;
 extern int remove_mapping(struct address_space *mapping, struct page *page);
 extern unsigned long vm_total_pages;
 
+extern unsigned long reclaim_pages(struct list_head *page_list);
 #ifdef CONFIG_NUMA
 extern int node_reclaim_mode;
 extern int sysctl_min_unmapped_ratio;
include/uapi/asm-generic/mman-common.h +1 -0
index d7b4231eea63..f545e159b472 100644
@@ -48,6 +48,7 @@
 #define MADV_WILLNEED	3		/* will need these pages */
 #define MADV_DONTNEED	4		/* don't need these pages */
 #define MADV_COLD	5		/* deactivatie these pages */
+#define MADV_PAGEOUT	6		/* reclaim these pages */
 
 /* common parameters: try to keep these consistent across architectures */
 #define MADV_FREE	8		/* free pages only if memory pressure */
mm/madvise.c +212 -0

242 lines not on this page · raw message

mm/vmscan.c +58 -0

64 lines not on this page · raw message

Signature · 1 line
2.22.0.410.gd8fdbe21b5-goog
#20 lore raw

PATCH v3 5/5 mm: factor out pmd young/dirty bit handling and THP split

In reply to Minchan Kim

Now, there are common part among MADV_COLD|PAGEOUT|FREE to reset access/dirty bit resetting or split the THP page to handle part of subpages in the THP page. This patch factor out the common part.

---
 include/linux/huge_mm.h |   3 -
 mm/huge_memory.c        |  74 -------------
 mm/madvise.c            | 234 +++++++++++++++++++++++-----------------
 3 files changed, 135 insertions(+), 176 deletions(-)
include/linux/huge_mm.h +0 -3
index 7cd5c150c21d..2667e1aa3ce5 100644
@@ -29,9 +29,6 @@ extern struct page *follow_trans_huge_pmd(struct vm_area_struct *vma,
 					  unsigned long addr,
 					  pmd_t *pmd,
 					  unsigned int flags);
-extern bool madvise_free_huge_pmd(struct mmu_gather *tlb,
-			struct vm_area_struct *vma,
-			pmd_t *pmd, unsigned long addr, unsigned long next);
 extern int zap_huge_pmd(struct mmu_gather *tlb,
 			struct vm_area_struct *vma,
 			pmd_t *pmd, unsigned long addr);
mm/huge_memory.c +0 -74

80 lines not on this page · raw message

mm/madvise.c +135 -99

307 lines not on this page · raw message

Signature · 1 line
2.22.0.410.gd8fdbe21b5-goog
#2 lore raw

Re: [PATCH v3 1/5] mm: introduce MADV_COLD

In reply to Minchan Kim
Quoted 9 lines · Minchan Kim

This patch introduces the new MADV_COLD hint to madvise(2) syscall. MADV_COLD can be used by a process to mark a memory range as not expected to be used in the near future. The hint can help kernel in deciding which pages to evict early during memory pressure.

It works for every LRU pages like MADV_[DONTNEED|FREE]. IOW, It moves

	active file page -> inactive file LRU
	active anon page -> inacdtive anon LRU

Is the LRU behavior part of the interface or the implementation?

I ask because we've got something in between tossing something down the LRU and swapping it: page migration. Specifically, on a system with slower memory media (like persistent memory) we just migrate a page instead of discarding it at reclaim:

Quoted 1 line

https://lore.kernel.org/linux-mm/[email protected]/

So let's say I have a page I want to evict from DRAM to the next slower tier of memory. Do I use MADV_COLD or MADV_PAGEOUT? If the LRU behavior is part of the interface itself, then MADV_COLD doesn't work.

Do you think we'll need a third MADV_ flag for our automatic migration behavior? MADV_REALLYCOLD? MADV_MIGRATEOUT?

reviewer
#3 lore raw

Re: [PATCH v3 1/5] mm: introduce MADV_COLD

Quoted 11 lines · Dave Hansen
Quoted 9 lines · Minchan Kim

This patch introduces the new MADV_COLD hint to madvise(2) syscall. MADV_COLD can be used by a process to mark a memory range as not expected to be used in the near future. The hint can help kernel in deciding which pages to evict early during memory pressure.

It works for every LRU pages like MADV_[DONTNEED|FREE]. IOW, It moves

	active file page -> inactive file LRU
	active anon page -> inacdtive anon LRU
Quoted 6 lines

Is the LRU behavior part of the interface or the implementation?

I ask because we've got something in between tossing something down the LRU and swapping it: page migration. Specifically, on a system with slower memory media (like persistent memory) we just migrate a page instead of discarding it at reclaim:

But we already do have interfaces for migrating the memory (move_pages(2)). Why should this interface duplicate that interface? I believe the only purpose of these two new madvise modes is to provide a non-destructive MADV_{DONTNEED,FREE} alteternatives. In other words, pageout vs. age interface.

Signature · 2 lines
Michal Hocko
SUSE Labs
#4 lore raw

Re: [PATCH v3 1/5] mm: introduce MADV_COLD

Quoted 11 lines · Michal Hocko
Quoted 6 lines

Is the LRU behavior part of the interface or the implementation?

I ask because we've got something in between tossing something down the LRU and swapping it: page migration. Specifically, on a system with slower memory media (like persistent memory) we just migrate a page instead of discarding it at reclaim:

But we already do have interfaces for migrating the memory (move_pages(2)). Why should this interface duplicate that interface? I believe the only purpose of these two new madvise modes is to provide a non-destructive MADV_{DONTNEED,FREE} alteternatives. In other words, pageout vs. age interface.

The existing interface's problem for this case is that it has to know exact locations where the memory is and where it should go. For instance, if you have two sockets, you very likely want to demote DRAM to the persistent memory DIMM sitting next to it and not go cross-socket. To do _that_, you need to know where the existing allocation lies so you can find the appropriate destination node.

That's not a problem for existing NUMA-enlightened apps, but it is for everything else.

For MADV_COLD, if we defined it like this, I think we could use it for both purposes (demotion and LRU movement):

Pages in the specified regions will be treated as less-recently- accessed compared to pages in the system with similar access frequencies. In contrast to MADV_DONTNEED, the contents of the region are preserved.

It would be nice not to talk about reclaim at all since we're not promising reclaim per se.

reviewer
#5 lore raw

Re: [PATCH v3 1/5] mm: introduce MADV_COLD

On Thu 27-06-19 07:36:50, Dave Hansen wrote:
[...]

Quoted 6 lines

For MADV_COLD, if we defined it like this, I think we could use it for both purposes (demotion and LRU movement):

Pages in the specified regions will be treated as less-recently- accessed compared to pages in the system with similar access frequencies. In contrast to MADV_DONTNEED, the contents of the

you meant s@MADV_DONTNEED@MADV_FREE@ I suppose

Quoted 4 lines

region are preserved.

It would be nice not to talk about reclaim at all since we're not promising reclaim per se.

Well, I guess this is just an implementation detail. MADV_FREE is really only about aging. It is up to the kernel what to do during the reclaim and the advice doesn't and shouldn't make any difference here.

Now MADV_PAGEOUT would be more tricky in that direction because it defines an immediate action to page out the range. I do understand your argument about NUMA unaware applications which might want to get something like MADV_DEMOTE which would move a page to a secondary memory (whatever that is) but I think this is asking for its own madvise. MADV_PAGEOUT has a quite simple semnatic - move to the backing storage - and I would rather not make it more complex.

Signature · 2 lines
Michal Hocko
SUSE Labs
#23 lore raw
In reply to Minchan Kim
Quoted 12 lines · Minchan Kim

- Problem

Naturally, cached apps were dominant consumers of memory on the system. However, they were not significant consumers of swap even though they are good candidate for swap. Under investigation, swapping out only begins once the low zone watermark is hit and kswapd wakes up, but the overall allocation rate in the system might trip lmkd thresholds and cause a cached process to be killed(we measured performance swapping out vs. zapping the memory by killing a process. Unsurprisingly, zapping is 10x times faster even though we use zram which is much faster than real storage) so kill from lmkd will often satisfy the high zone watermark, resulting in very few pages actually being moved to swap.

Maybe we should look if we do The Right Thing™ at system-wide level before introducing new API? How changing swappiness affects your workloads? What is swappiness value in your setup?

Signature · 1 line
 Kirill A. Shutemov
#24 lore raw
Quoted 14 lines · Kirill A. Shutemov
Quoted 12 lines · Minchan Kim

- Problem

Naturally, cached apps were dominant consumers of memory on the system. However, they were not significant consumers of swap even though they are good candidate for swap. Under investigation, swapping out only begins once the low zone watermark is hit and kswapd wakes up, but the overall allocation rate in the system might trip lmkd thresholds and cause a cached process to be killed(we measured performance swapping out vs. zapping the memory by killing a process. Unsurprisingly, zapping is 10x times faster even though we use zram which is much faster than real storage) so kill from lmkd will often satisfy the high zone watermark, resulting in very few pages actually being moved to swap.

Quoted 3 lines

Maybe we should look if we do The Right Thing™ at system-wide level before introducing new API? How changing swappiness affects your workloads? What is swappiness value in your setup?

It was 100. Even, I tried 150 and 200 with simple hack of swappiness. However, it caused too excessive swpout.

Anyway, systen-level tune is generally good but if process has hint, that should work better and that's why advise API is.

#9 lore raw

Re: [PATCH v3 1/5] mm: introduce MADV_COLD

In reply to Dave Hansen
Quoted 6 lines · Dave Hansen
Quoted 5 lines · Minchan Kim

This patch introduces the new MADV_COLD hint to madvise(2) syscall. MADV_COLD can be used by a process to mark a memory range as not expected to be used in the near future. The hint can help kernel in deciding which pages to evict early during memory pressure.

Quoted 6 lines
Quoted 4 lines

It works for every LRU pages like MADV_[DONTNEED|FREE]. IOW, It moves

	active file page -> inactive file LRU
	active anon page -> inacdtive anon LRU

Is the LRU behavior part of the interface or the implementation?

It's a just implementation. What user should expect with this API is they just informs to the kernel "this memory in the regions wouldn't access in the near future" so how kernel will handle memory in there is up to the kernel.

Quoted 11 lines

I ask because we've got something in between tossing something down the LRU and swapping it: page migration. Specifically, on a system with slower memory media (like persistent memory) we just migrate a page instead of discarding it at reclaim:

Quoted 1 line

https://lore.kernel.org/linux-mm/[email protected]/

So let's say I have a page I want to evict from DRAM to the next slower tier of memory. Do I use MADV_COLD or MADV_PAGEOUT? If the LRU behavior is part of the interface itself, then MADV_COLD doesn't work.

IMHO, if it's one of storage in the memory hierarchy, that shouldn't be transparent for the user? What I meant is VM moves inactive pages to the persistent memory before the reclaiming. IOW, VM would have one more level LRU or extened inactive LRU to cover the persistent memory.

Quoted 3 lines

Do you think we'll need a third MADV_ flag for our automatic migration behavior? MADV_REALLYCOLD? MADV_MIGRATEOUT?

I believe it depends on how we abstract the persistent memory of cache hierarchy. If we abstract it as diffrent storage with DRAM, manybe, that should be part of other syscall like like move_pages.
If we abstract it as part of DRAM, that should be part of additional LRU or extended inactive LRU.

#6 lore raw

Re: [PATCH v3 1/5] mm: introduce MADV_COLD

In reply to Michal Hocko
Quoted 10 lines · Michal Hocko

On Thu 27-06-19 07:36:50, Dave Hansen wrote:
[...]

Quoted 6 lines

For MADV_COLD, if we defined it like this, I think we could use it for both purposes (demotion and LRU movement):

Pages in the specified regions will be treated as less-recently- accessed compared to pages in the system with similar access frequencies. In contrast to MADV_DONTNEED, the contents of the

you meant s@MADV_DONTNEED@MADV_FREE@ I suppose

Right, MADV_FREE is more proper because it's aging related.

Quoted 5 lines
Quoted 5 lines

region are preserved.

It would be nice not to talk about reclaim at all since we're not promising reclaim per se.

Your suggestion doesn't expose any implementation detail and could meet your needs later. I'm okay. I will change it if others are not against of it.

Thanks, Dave.

#7 lore raw

Re: [PATCH v3 1/5] mm: introduce MADV_COLD

Quoted 20 lines · Minchan Kim
Quoted 10 lines · Michal Hocko

On Thu 27-06-19 07:36:50, Dave Hansen wrote:
[...]

Quoted 6 lines

For MADV_COLD, if we defined it like this, I think we could use it for both purposes (demotion and LRU movement):

Pages in the specified regions will be treated as less-recently- accessed compared to pages in the system with similar access frequencies. In contrast to MADV_DONTNEED, the contents of the

you meant s@MADV_DONTNEED@MADV_FREE@ I suppose

Right, MADV_FREE is more proper because it's aging related.

Quoted 5 lines
Quoted 5 lines

region are preserved.

It would be nice not to talk about reclaim at all since we're not promising reclaim per se.

Quoted 6 lines

Your suggestion doesn't expose any implementation detail and could meet your needs later. I'm okay. I will change it if others are not against of it.

Thanks, Dave.

From 39df9f94e6204b8893f3f3feb692745657392657 Mon Sep 17 00:00:00 2001

From: Minchan Kim <[email protected]>
Date: Fri, 24 May 2019 13:47:54 +0900
Subject: [PATCH v3 1/5] mm: introduce MADV_COLD

When a process expects no accesses to a certain memory range, it could give a hint to kernel that the pages can be reclaimed when memory pressure happens but data should be preserved for future use. This could reduce workingset eviction so it ends up increasing performance.

This patch introduces the new MADV_COLD hint to madvise(2) syscall. MADV_COLD can be used by a process to mark a memory range as not expected to be used in the near future. The hint can help kernel in deciding which pages to evict early during memory pressure.

It works for every LRU pages like MADV_[DONTNEED|FREE]. IOW, It moves

	active file page -> inactive file LRU
	active anon page -> inacdtive anon LRU

Unlike MADV_FREE, it doesn't move active anonymous pages to inactive file LRU's head because MADV_COLD is a little bit different symantic. MADV_FREE means it's okay to discard when the memory pressure because the content of the page is *garbage* so freeing such pages is almost zero overhead since we don't need to swap out and access afterward causes just minor fault. Thus, it would make sense to put those freeable pages in inactive file LRU to compete other used-once pages. It makes sense for implmentaion point of view, too because it's not swapbacked memory any longer until it would be re-dirtied. Even, it could give a bonus to make them be reclaimed on swapless system. However, MADV_COLD doesn't mean garbage so reclaiming them requires swap-out/in in the end so it's bigger cost. Since we have designed VM LRU aging based on cost-model, anonymous cold pages would be better to position inactive anon's LRU list, not file LRU. Furthermore, it would help to avoid unnecessary scanning if system doesn't have a swap device. Let's start simpler way without adding complexity at this moment. However, keep in mind, too that it's a caveat that workloads with a lot of pages cache are likely to ignore MADV_COLD on anonymous memory because we rarely age anonymous LRU lists.

* man-page material

MADV_COLD (since Linux x.x)

Pages in the specified regions will be treated as less-recently-accessed compared to pages in the system with similar access frequencies. In contrast to MADV_FREE, the contents of the region are preserved regardless of subsequent writes to pages.

MADV_COLD cannot be applied to locked pages, Huge TLB pages, or VM_PFNMAP pages.

* v2
* add up the warn with lots of page cache workload - mhocko
* add man page stuff - dave

* v1
* remove page_mapcount filter - hannes, mhocko
* remove idle page handling - joelaf

* RFCv2
* add more description - mhocko

* RFCv1
* renaming from MADV_COOL to MADV_COLD - hannes

* internal review
* use clear_page_youn in deactivate_page - joelaf
* Revise the description - surenb
* Renaming from MADV_WARM to MADV_COOL - surenb

---
 include/linux/swap.h                   |   1 +
 include/uapi/asm-generic/mman-common.h |   1 +
 mm/internal.h                          |   2 +-
 mm/madvise.c                           | 180 ++++++++++++++++++++++++-
 mm/oom_kill.c                          |   2 +-
 mm/swap.c                              |  42 ++++++
 6 files changed, 224 insertions(+), 4 deletions(-)
include/linux/swap.h +1 -0
index de2c67a33b7e..0ce997edb8bb 100644
@@ -340,6 +340,7 @@ extern void lru_add_drain_cpu(int cpu);
 extern void lru_add_drain_all(void);
 extern void rotate_reclaimable_page(struct page *page);
 extern void deactivate_file_page(struct page *page);
+extern void deactivate_page(struct page *page);
 extern void mark_page_lazyfree(struct page *page);
 extern void swap_setup(void);
 
include/uapi/asm-generic/mman-common.h +1 -0
index ef4623f03156..d7b4231eea63 100644
@@ -47,6 +47,7 @@
 #define MADV_SEQUENTIAL	2		/* expect sequential page references */
 #define MADV_WILLNEED	3		/* will need these pages */
 #define MADV_DONTNEED	4		/* don't need these pages */
+#define MADV_COLD	5		/* deactivatie these pages */
 
 /* common parameters: try to keep these consistent across architectures */
 #define MADV_FREE	8		/* free pages only if memory pressure */
mm/internal.h +1 -1
index f53a14d67538..c61b215ff265 100644
@@ -39,7 +39,7 @@ vm_fault_t do_swap_page(struct vm_fault *vmf);
 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
 		unsigned long floor, unsigned long ceiling);
 
-static inline bool can_madv_dontneed_vma(struct vm_area_struct *vma)
+static inline bool can_madv_lru_vma(struct vm_area_struct *vma)
 {
 	return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP));
 }
mm/madvise.c +178 -2

216 lines not on this page · raw message

mm/oom_kill.c +1 -1
index 6de5c354d6ca..2140a6f8db63 100644
@@ -523,7 +523,7 @@ bool __oom_reap_task_mm(struct mm_struct *mm)
 	set_bit(MMF_UNSTABLE, &mm->flags);
 
 	for (vma = mm->mmap ; vma; vma = vma->vm_next) {
-		if (!can_madv_dontneed_vma(vma))
+		if (!can_madv_lru_vma(vma))
 			continue;
 
 		/*
mm/swap.c +42 -0

72 lines not on this page · raw message

Signature · 1 line
2.22.0.410.gd8fdbe21b5-goog
#25 lore raw
In reply to Minchan Kim

Hi Folks,

Do you guys have comments? I think it would be long enough to be pending. If there is no further comments, I want to ask to merge.

Thanks.

Quoted 99 lines · Minchan Kim

This patch is part of previous series:
https://lore.kernel.org/lkml/[email protected]/T/#u
Originally, it was created for external madvise hinting feature.

https://lkml.org/lkml/2019/5/31/463
Michal wanted to separte the discussion from external hinting interface so this patchset includes only first part of my entire patchset

- introduce MADV_COLD and MADV_PAGEOUT hint to madvise.

However, I keep entire description for others for easier understanding why this kinds of hint was born.

Thanks.

This patchset is against on next-20190530.

Below is description of previous entire patchset.
================= &< =====================

- Background

The Android terminology used for forking a new process and starting an app from scratch is a cold start, while resuming an existing app is a hot start. While we continually try to improve the performance of cold starts, hot starts will always be significantly less power hungry as well as faster so we are trying to make hot start more likely than cold start.

To increase hot start, Android userspace manages the order that apps should be killed in a process called ActivityManagerService. ActivityManagerService tracks every Android app or service that the user could be interacting with at any time and translates that into a ranked list for lmkd(low memory killer daemon). They are likely to be killed by lmkd if the system has to reclaim memory. In that sense they are similar to entries in any other cache. Those apps are kept alive for opportunistic performance improvements but those performance improvements will vary based on the memory requirements of individual workloads.

- Problem

Naturally, cached apps were dominant consumers of memory on the system. However, they were not significant consumers of swap even though they are good candidate for swap. Under investigation, swapping out only begins once the low zone watermark is hit and kswapd wakes up, but the overall allocation rate in the system might trip lmkd thresholds and cause a cached process to be killed(we measured performance swapping out vs. zapping the memory by killing a process. Unsurprisingly, zapping is 10x times faster even though we use zram which is much faster than real storage) so kill from lmkd will often satisfy the high zone watermark, resulting in very few pages actually being moved to swap.

- Approach

The approach we chose was to use a new interface to allow userspace to proactively reclaim entire processes by leveraging platform information. This allowed us to bypass the inaccuracy of the kernel’s LRUs for pages that are known to be cold from userspace and to avoid races with lmkd by reclaiming apps as soon as they entered the cached state. Additionally, it could provide many chances for platform to use much information to optimize memory efficiency.

To achieve the goal, the patchset introduce two new options for madvise. One is MADV_COLD which will deactivate activated pages and the other is MADV_PAGEOUT which will reclaim private pages instantly. These new options complement MADV_DONTNEED and MADV_FREE by adding non-destructive ways to gain some free memory space. MADV_PAGEOUT is similar to MADV_DONTNEED in a way that it hints the kernel that memory region is not currently needed and should be reclaimed immediately; MADV_COLD is similar to MADV_FREE in a way that it hints the kernel that memory region is not currently needed and should be reclaimed when memory pressure rises.

Minchan Kim (5):
mm: introduce MADV_COLD
mm: change PAGEREF_RECLAIM_CLEAN with PAGE_REFRECLAIM
mm: account nr_isolated_xxx in [isolate|putback]_lru_page
mm: introduce MADV_PAGEOUT
mm: factor out pmd young/dirty bit handling and THP split

 include/linux/huge_mm.h                |   3 -
 include/linux/swap.h                   |   2 +
 include/uapi/asm-generic/mman-common.h |   2 +
 mm/compaction.c                        |   2 -
 mm/gup.c                               |   7 +-
 mm/huge_memory.c                       |  74 -----
 mm/internal.h                          |   2 +-
 mm/khugepaged.c                        |   3 -
 mm/madvise.c                           | 438 ++++++++++++++++++++++++-
 mm/memory-failure.c                    |   3 -
 mm/memory_hotplug.c                    |   4 -
 mm/mempolicy.c                         |   6 +-
 mm/migrate.c                           |  37 +--
 mm/oom_kill.c                          |   2 +-
 mm/swap.c                              |  42 +++
 mm/vmscan.c                            |  86 ++++-
 16 files changed, 566 insertions(+), 147 deletions(-)
Signature · 1 line
2.22.0.410.gd8fdbe21b5-goog
reviewer
#26 lore raw
Quoted 5 lines · Minchan Kim

Hi Folks,

Do you guys have comments? I think it would be long enough to be pending. If there is no further comments, I want to ask to merge.

This is definitely on my todo list for this week. But please be patient. It's been _one_ work day since you posted this last version so I do not think this is stalling for too long. Sure the current version is probably not too much different from the previous but I didn't get to review it in the depth yet. All the code duplication doesn't make it much easier but I understand your reasoning that sharing more code is not really straightforward.

Signature · 2 lines
Michal Hocko
SUSE Labs
reviewer Acked-by
#8 lore raw

Re: [PATCH v3 1/5] mm: introduce MADV_COLD

In reply to Minchan Kim
Quoted 65 lines · Minchan Kim
Quoted 1 line

From 39df9f94e6204b8893f3f3feb692745657392657 Mon Sep 17 00:00:00 2001

From: Minchan Kim <[email protected]>
Date: Fri, 24 May 2019 13:47:54 +0900
Subject: [PATCH v3 1/5] mm: introduce MADV_COLD

When a process expects no accesses to a certain memory range, it could give a hint to kernel that the pages can be reclaimed when memory pressure happens but data should be preserved for future use. This could reduce workingset eviction so it ends up increasing performance.

This patch introduces the new MADV_COLD hint to madvise(2) syscall. MADV_COLD can be used by a process to mark a memory range as not expected to be used in the near future. The hint can help kernel in deciding which pages to evict early during memory pressure.

It works for every LRU pages like MADV_[DONTNEED|FREE]. IOW, It moves

	active file page -> inactive file LRU
	active anon page -> inacdtive anon LRU

Unlike MADV_FREE, it doesn't move active anonymous pages to inactive file LRU's head because MADV_COLD is a little bit different symantic. MADV_FREE means it's okay to discard when the memory pressure because the content of the page is *garbage* so freeing such pages is almost zero overhead since we don't need to swap out and access afterward causes just minor fault. Thus, it would make sense to put those freeable pages in inactive file LRU to compete other used-once pages. It makes sense for implmentaion point of view, too because it's not swapbacked memory any longer until it would be re-dirtied. Even, it could give a bonus to make them be reclaimed on swapless system. However, MADV_COLD doesn't mean garbage so reclaiming them requires swap-out/in in the end so it's bigger cost. Since we have designed VM LRU aging based on cost-model, anonymous cold pages would be better to position inactive anon's LRU list, not file LRU. Furthermore, it would help to avoid unnecessary scanning if system doesn't have a swap device. Let's start simpler way without adding complexity at this moment. However, keep in mind, too that it's a caveat that workloads with a lot of pages cache are likely to ignore MADV_COLD on anonymous memory because we rarely age anonymous LRU lists.

* man-page material

MADV_COLD (since Linux x.x)

Pages in the specified regions will be treated as less-recently-accessed compared to pages in the system with similar access frequencies. In contrast to MADV_FREE, the contents of the region are preserved regardless of subsequent writes to pages.

MADV_COLD cannot be applied to locked pages, Huge TLB pages, or VM_PFNMAP pages.

* v2
* add up the warn with lots of page cache workload - mhocko
* add man page stuff - dave

* v1
* remove page_mapcount filter - hannes, mhocko
* remove idle page handling - joelaf

* RFCv2
* add more description - mhocko

* RFCv1
* renaming from MADV_COOL to MADV_COLD - hannes

Quoted 6 lines

* internal review
* use clear_page_youn in deactivate_page - joelaf
* Revise the description - surenb
* Renaming from MADV_WARM to MADV_COOL - surenb

OK, looks reasonable to me. THP part still gives me a head spin but it is consistent with madv_free part so I will trust that all weird corner cases are already caught there.

Thanks!

Signature · 2 lines
Michal Hocko
SUSE Labs
reviewer Acked-by on v3 3/5
#12 lore raw

Re: [PATCH v3 3/5] mm: account nr_isolated_xxx in [isolate|putback]_lru_page

In reply to Minchan Kim
Quoted 11 lines · Minchan Kim

The isolate counting is pecpu counter so it would be not huge gain to work them by batch. Rather than complicating to make them batch, let's make it more stright-foward via adding the counting logic into [isolate|putback]_lru_page API.

* v1
* fix accounting bug - Hillf

I like that the NR_ISOLATED_$FOO handling gets out of any code except for vmscan and migration. This is definitely an improvement.

I haven't spotted any imbalance so I hope I haven't really missed any path.

Thanks!

Quoted 330 lines
---
 mm/compaction.c     |  2 --
 mm/gup.c            |  7 +------
 mm/khugepaged.c     |  3 ---
 mm/memory-failure.c |  3 ---
 mm/memory_hotplug.c |  4 ----
 mm/mempolicy.c      |  6 +-----
 mm/migrate.c        | 37 ++++++++-----------------------------
 mm/vmscan.c         | 22 ++++++++++++++++------
 8 files changed, 26 insertions(+), 58 deletions(-)
mm/compaction.c +0 -2
index 9e1b9acb116b..c6591682deda 100644
@@ -982,8 +982,6 @@ isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn,
 
 		/* Successfully isolated */
 		del_page_from_lru_list(page, lruvec, page_lru(page));
-		inc_node_page_state(page,
-				NR_ISOLATED_ANON + page_is_file_cache(page));
 
 isolate_success:
 		list_add(&page->lru, &cc->migratepages);
mm/gup.c +1 -6
index 7dde2e3a1963..aec3a2b7e61b 100644
@@ -1473,13 +1473,8 @@ static long check_and_migrate_cma_pages(struct task_struct *tsk,
 					drain_allow = false;
 				}
 
-				if (!isolate_lru_page(head)) {
+				if (!isolate_lru_page(head))
 					list_add_tail(&head->lru, &cma_page_list);
-					mod_node_page_state(page_pgdat(head),
-							    NR_ISOLATED_ANON +
-							    page_is_file_cache(head),
-							    hpage_nr_pages(head));
-				}
 			}
 		}
 	}
mm/khugepaged.c +0 -3
index 0f7419938008..7da34e198ec5 100644
@@ -503,7 +503,6 @@ void __khugepaged_exit(struct mm_struct *mm)
 
 static void release_pte_page(struct page *page)
 {
-	dec_node_page_state(page, NR_ISOLATED_ANON + page_is_file_cache(page));
 	unlock_page(page);
 	putback_lru_page(page);
 }
@@ -602,8 +601,6 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 			result = SCAN_DEL_PAGE_LRU;
 			goto out;
 		}
-		inc_node_page_state(page,
-				NR_ISOLATED_ANON + page_is_file_cache(page));
 		VM_BUG_ON_PAGE(!PageLocked(page), page);
 		VM_BUG_ON_PAGE(PageLRU(page), page);
 
mm/memory-failure.c +0 -3
index 7e08cbf3ba49..3586e8226e4e 100644
@@ -1795,9 +1795,6 @@ static int __soft_offline_page(struct page *page, int flags)
 		 * so use !__PageMovable instead for LRU page's mapping
 		 * cannot have PAGE_MAPPING_MOVABLE.
 		 */
-		if (!__PageMovable(page))
-			inc_node_page_state(page, NR_ISOLATED_ANON +
-						page_is_file_cache(page));
 		list_add(&page->lru, &pagelist);
 		ret = migrate_pages(&pagelist, new_page, NULL, MPOL_MF_MOVE_ALL,
 					MIGRATE_SYNC, MR_MEMORY_FAILURE);
mm/memory_hotplug.c +0 -4
index dfab21dc33dc..68577c677b46 100644
@@ -1384,10 +1384,6 @@ do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
 			ret = isolate_movable_page(page, ISOLATE_UNEVICTABLE);
 		if (!ret) { /* Success */
 			list_add_tail(&page->lru, &source);
-			if (!__PageMovable(page))
-				inc_node_page_state(page, NR_ISOLATED_ANON +
-						    page_is_file_cache(page));
-
 		} else {
 			pr_warn("failed to isolate pfn %lx\n", pfn);
 			dump_page(page, "isolation failed");
mm/mempolicy.c +1 -5
index 64562809bf3b..03081f3404ca 100644
@@ -994,12 +994,8 @@ static int migrate_page_add(struct page *page, struct list_head *pagelist,
 	 * Avoid migrating a page that is shared with others.
 	 */
 	if ((flags & MPOL_MF_MOVE_ALL) || page_mapcount(head) == 1) {
-		if (!isolate_lru_page(head)) {
+		if (!isolate_lru_page(head))
 			list_add_tail(&head->lru, pagelist);
-			mod_node_page_state(page_pgdat(head),
-				NR_ISOLATED_ANON + page_is_file_cache(head),
-				hpage_nr_pages(head));
-		}
 	}
 
 	return 0;
mm/migrate.c +8 -29
index 572b4bc85d76..5583324c01e7 100644
@@ -190,8 +190,6 @@ void putback_movable_pages(struct list_head *l)
 			unlock_page(page);
 			put_page(page);
 		} else {
-			mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON +
-					page_is_file_cache(page), -hpage_nr_pages(page));
 			putback_lru_page(page);
 		}
 	}
@@ -1181,10 +1179,17 @@ static ICE_noinline int unmap_and_move(new_page_t get_new_page,
 		return -ENOMEM;
 
 	if (page_count(page) == 1) {
+		bool is_lru = !__PageMovable(page);
+
 		/* page was freed from under us. So we are done. */
 		ClearPageActive(page);
 		ClearPageUnevictable(page);
-		if (unlikely(__PageMovable(page))) {
+		if (likely(is_lru))
+			mod_node_page_state(page_pgdat(page),
+						NR_ISOLATED_ANON +
+						page_is_file_cache(page),
+						-hpage_nr_pages(page));
+		else {
 			lock_page(page);
 			if (!PageMovable(page))
 				__ClearPageIsolated(page);
@@ -1210,15 +1215,6 @@ static ICE_noinline int unmap_and_move(new_page_t get_new_page,
 		 * restored.
 		 */
 		list_del(&page->lru);
-
-		/*
-		 * Compaction can migrate also non-LRU pages which are
-		 * not accounted to NR_ISOLATED_*. They can be recognized
-		 * as __PageMovable
-		 */
-		if (likely(!__PageMovable(page)))
-			mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON +
-					page_is_file_cache(page), -hpage_nr_pages(page));
 	}
 
 	/*
@@ -1572,9 +1568,6 @@ static int add_page_for_migration(struct mm_struct *mm, unsigned long addr,
 
 		err = 0;
 		list_add_tail(&head->lru, pagelist);
-		mod_node_page_state(page_pgdat(head),
-			NR_ISOLATED_ANON + page_is_file_cache(head),
-			hpage_nr_pages(head));
 	}
 out_putpage:
 	/*
@@ -1890,8 +1883,6 @@ static struct page *alloc_misplaced_dst_page(struct page *page,
 
 static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
 {
-	int page_lru;
-
 	VM_BUG_ON_PAGE(compound_order(page) && !PageTransHuge(page), page);
 
 	/* Avoid migrating to a node that is nearly full */
@@ -1913,10 +1904,6 @@ static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
 		return 0;
 	}
 
-	page_lru = page_is_file_cache(page);
-	mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON + page_lru,
-				hpage_nr_pages(page));
-
 	/*
 	 * Isolating the page has taken another reference, so the
 	 * caller's reference can be safely dropped without the page
@@ -1971,8 +1958,6 @@ int migrate_misplaced_page(struct page *page, struct vm_area_struct *vma,
 	if (nr_remaining) {
 		if (!list_empty(&migratepages)) {
 			list_del(&page->lru);
-			dec_node_page_state(page, NR_ISOLATED_ANON +
-					page_is_file_cache(page));
 			putback_lru_page(page);
 		}
 		isolated = 0;
@@ -2002,7 +1987,6 @@ int migrate_misplaced_transhuge_page(struct mm_struct *mm,
 	pg_data_t *pgdat = NODE_DATA(node);
 	int isolated = 0;
 	struct page *new_page = NULL;
-	int page_lru = page_is_file_cache(page);
 	unsigned long start = address & HPAGE_PMD_MASK;
 
 	new_page = alloc_pages_node(node,
@@ -2048,8 +2032,6 @@ int migrate_misplaced_transhuge_page(struct mm_struct *mm,
 		/* Retake the callers reference and putback on LRU */
 		get_page(page);
 		putback_lru_page(page);
-		mod_node_page_state(page_pgdat(page),
-			 NR_ISOLATED_ANON + page_lru, -HPAGE_PMD_NR);
 
 		goto out_unlock;
 	}
@@ -2099,9 +2081,6 @@ int migrate_misplaced_transhuge_page(struct mm_struct *mm,
 	count_vm_events(PGMIGRATE_SUCCESS, HPAGE_PMD_NR);
 	count_vm_numa_events(NUMA_PAGE_MIGRATE, HPAGE_PMD_NR);
 
-	mod_node_page_state(page_pgdat(page),
-			NR_ISOLATED_ANON + page_lru,
-			-HPAGE_PMD_NR);
 	return isolated;
 
 out_fail:
mm/vmscan.c +16 -6
index 49e9ee4d771d..223ce5da08f0 100644
@@ -1014,6 +1014,9 @@ int remove_mapping(struct address_space *mapping, struct page *page)
 void putback_lru_page(struct page *page)
 {
 	lru_cache_add(page);
+	mod_node_page_state(page_pgdat(page),
+				NR_ISOLATED_ANON + page_is_file_cache(page),
+				-hpage_nr_pages(page));
 	put_page(page);		/* drop ref from isolate */
 }
 
@@ -1479,6 +1482,9 @@ static unsigned long shrink_page_list(struct list_head *page_list,
 		 */
 		nr_reclaimed += nr_pages;
 
+		mod_node_page_state(pgdat, NR_ISOLATED_ANON +
+						page_is_file_cache(page),
+						-nr_pages);
 		/*
 		 * Is there need to periodically free_page_list? It would
 		 * appear not as the counts should be low
@@ -1554,7 +1560,6 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 	ret = shrink_page_list(&clean_pages, zone->zone_pgdat, &sc,
 			TTU_IGNORE_ACCESS, &dummy_stat, true);
 	list_splice(&clean_pages, page_list);
-	mod_node_page_state(zone->zone_pgdat, NR_ISOLATED_FILE, -ret);
 	return ret;
 }
 
@@ -1630,6 +1635,9 @@ int __isolate_lru_page(struct page *page, isolate_mode_t mode)
 		 */
 		ClearPageLRU(page);
 		ret = 0;
+		__mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON +
+						page_is_file_cache(page),
+						hpage_nr_pages(page));
 	}
 
 	return ret;
@@ -1761,6 +1769,7 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
 	trace_mm_vmscan_lru_isolate(sc->reclaim_idx, sc->order, nr_to_scan,
 				    total_scan, skipped, nr_taken, mode, lru);
 	update_lru_sizes(lruvec, lru, nr_zone_taken);
+
 	return nr_taken;
 }
 
@@ -1809,6 +1818,9 @@ int isolate_lru_page(struct page *page)
 			ClearPageLRU(page);
 			del_page_from_lru_list(page, lruvec, lru);
 			ret = 0;
+			mod_node_page_state(pgdat, NR_ISOLATED_ANON +
+						page_is_file_cache(page),
+						hpage_nr_pages(page));
 		}
 		spin_unlock_irq(&pgdat->lru_lock);
 	}
@@ -1900,6 +1912,9 @@ static unsigned noinline_for_stack move_pages_to_lru(struct lruvec *lruvec,
 		update_lru_size(lruvec, lru, page_zonenum(page), nr_pages);
 		list_move(&page->lru, &lruvec->lists[lru]);
 
+		__mod_node_page_state(pgdat, NR_ISOLATED_ANON +
+						page_is_file_cache(page),
+						-hpage_nr_pages(page));
 		if (put_page_testzero(page)) {
 			__ClearPageLRU(page);
 			__ClearPageActive(page);
@@ -1977,7 +1992,6 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
 	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &page_list,
 				     &nr_scanned, sc, lru);
 
-	__mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, nr_taken);
 	reclaim_stat->recent_scanned[file] += nr_taken;
 
 	item = current_is_kswapd() ? PGSCAN_KSWAPD : PGSCAN_DIRECT;
@@ -2003,8 +2017,6 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
 
 	move_pages_to_lru(lruvec, &page_list);
 
-	__mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, -nr_taken);
-
 	spin_unlock_irq(&pgdat->lru_lock);
 
 	mem_cgroup_uncharge_list(&page_list);
@@ -2063,7 +2075,6 @@ static void shrink_active_list(unsigned long nr_to_scan,
 	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &l_hold,
 				     &nr_scanned, sc, lru);
 
-	__mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, nr_taken);
 	reclaim_stat->recent_scanned[file] += nr_taken;
 
 	__count_vm_events(PGREFILL, nr_scanned);
@@ -2132,7 +2143,6 @@ static void shrink_active_list(unsigned long nr_to_scan,
 	__count_vm_events(PGDEACTIVATE, nr_deactivate);
 	__count_memcg_events(lruvec_memcg(lruvec), PGDEACTIVATE, nr_deactivate);
 
-	__mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, -nr_taken);
 	spin_unlock_irq(&pgdat->lru_lock);
 
 	mem_cgroup_uncharge_list(&l_active);
Signature · 1 line
2.22.0.410.gd8fdbe21b5-goog
Signature · 2 lines
Michal Hocko
SUSE Labs
reviewer Acked-by on v3 4/5
#14 lore raw

Re: [PATCH v3 4/5] mm: introduce MADV_PAGEOUT

In reply to Minchan Kim
Quoted 22 lines · Minchan Kim

When a process expects no accesses to a certain memory range for a long time, it could hint kernel that the pages can be reclaimed instantly but data should be preserved for future use. This could reduce workingset eviction so it ends up increasing performance.

This patch introduces the new MADV_PAGEOUT hint to madvise(2) syscall. MADV_PAGEOUT can be used by a process to mark a memory range as not expected to be used for a long time so that kernel reclaims *any LRU* pages instantly. The hint can help kernel in deciding which pages to evict proactively.

- man-page material

MADV_PAGEOUT (since Linux x.x)

Do not expect access in the near future so pages in the specified regions could be reclaimed instantly regardless of memory pressure. Thus, access in the range after successful operation could cause major page fault but never lose the up-to-date contents unlike MADV_DONTNEED.

Quoted 4 lines

It works for only private anonymous mappings and non-anonymous mappings that belong to files that the calling process could successfully open for writing; otherwise, it could be used for sidechannel attack.

I would rephrase this way:
"
Pages belonging to a shared mapping are only processed if a write access is allowed for the calling process.
"

I wouldn't really mention side channel attacks for a man page. You can mention can_do_mincore check and the side channel prevention in the changelog that is not aimed for the man page.

Quoted 15 lines

MADV_PAGEOUT cannot be applied to locked pages, Huge TLB pages, or VM_PFNMAP pages.

* v2
* add comment about SWAP_CLUSTER_MAX - mhocko
* add permission check to prevent sidechannel attack - mhocko
* add man page stuff - dave

* v1
* change pte to old and rely on the other's reference - hannes
* remove page_mapcount to check shared page - mhocko

* RFC v2
* make reclaim_pages simple via factoring out isolate logic - hannes

Quoted 7 lines

* RFCv1
* rename from MADV_COLD to MADV_PAGEOUT - hannes
* bail out if process is being killed - Hillf
* fix reclaim_pages bugs - Hillf

---

I am still not convinced about the SWAP_CLUSTER_MAX batching and the udnerlying OOM argument. Is one pmd worth of pages really an OOM risk? Sure you can have many invocations in parallel and that would add on but the same might happen with SWAP_CLUSTER_MAX. So I would just remove the batching for now and think of it only if we really see this being a problem for real. Unless you feel really strong about this, of course.

Anyway the patch looks ok to me otherwise.

Signature · 2 lines
Michal Hocko
SUSE Labs
reviewer
#21 lore raw

Re: [PATCH v3 5/5] mm: factor out pmd young/dirty bit handling and THP split

In reply to Minchan Kim
Quoted 3 lines · Minchan Kim

Now, there are common part among MADV_COLD|PAGEOUT|FREE to reset access/dirty bit resetting or split the THP page to handle part of subpages in the THP page. This patch factor out the common part.

While this reduces the code duplication to some degree I suspect it only goes half way. I haven't tried that myself due to lack of time but I believe this has a potential to reduce even more. All those madvise calls are doing the same thing essentially. What page tables and apply an operation on ptes and/or a page that is mapped. And that suggests that the specific operation should be good with defining two - pte and page operations on each entry. All the rest should be a common code.

That being said, I do not feel strongly about this patch. The rest of the series should be good enough even without it and I wouldn't delay it just by discussing a potential of the cleanup.

Quoted 426 lines
---
 include/linux/huge_mm.h |   3 -
 mm/huge_memory.c        |  74 -------------
 mm/madvise.c            | 234 +++++++++++++++++++++++-----------------
 3 files changed, 135 insertions(+), 176 deletions(-)
include/linux/huge_mm.h +0 -3
index 7cd5c150c21d..2667e1aa3ce5 100644
@@ -29,9 +29,6 @@ extern struct page *follow_trans_huge_pmd(struct vm_area_struct *vma,
 					  unsigned long addr,
 					  pmd_t *pmd,
 					  unsigned int flags);
-extern bool madvise_free_huge_pmd(struct mmu_gather *tlb,
-			struct vm_area_struct *vma,
-			pmd_t *pmd, unsigned long addr, unsigned long next);
 extern int zap_huge_pmd(struct mmu_gather *tlb,
 			struct vm_area_struct *vma,
 			pmd_t *pmd, unsigned long addr);
mm/huge_memory.c +0 -74
index 93f531b63a45..e4b9a06788f3 100644
@@ -1671,80 +1671,6 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf, pmd_t pmd)
 	return 0;
 }
 
-/*
- * Return true if we do MADV_FREE successfully on entire pmd page.
- * Otherwise, return false.
- */
-bool madvise_free_huge_pmd(struct mmu_gather *tlb, struct vm_area_struct *vma,
-		pmd_t *pmd, unsigned long addr, unsigned long next)
-{
-	spinlock_t *ptl;
-	pmd_t orig_pmd;
-	struct page *page;
-	struct mm_struct *mm = tlb->mm;
-	bool ret = false;
-
-	tlb_change_page_size(tlb, HPAGE_PMD_SIZE);
-
-	ptl = pmd_trans_huge_lock(pmd, vma);
-	if (!ptl)
-		goto out_unlocked;
-
-	orig_pmd = *pmd;
-	if (is_huge_zero_pmd(orig_pmd))
-		goto out;
-
-	if (unlikely(!pmd_present(orig_pmd))) {
-		VM_BUG_ON(thp_migration_supported() &&
-				  !is_pmd_migration_entry(orig_pmd));
-		goto out;
-	}
-
-	page = pmd_page(orig_pmd);
-	/*
-	 * If other processes are mapping this page, we couldn't discard
-	 * the page unless they all do MADV_FREE so let's skip the page.
-	 */
-	if (page_mapcount(page) != 1)
-		goto out;
-
-	if (!trylock_page(page))
-		goto out;
-
-	/*
-	 * If user want to discard part-pages of THP, split it so MADV_FREE
-	 * will deactivate only them.
-	 */
-	if (next - addr != HPAGE_PMD_SIZE) {
-		get_page(page);
-		spin_unlock(ptl);
-		split_huge_page(page);
-		unlock_page(page);
-		put_page(page);
-		goto out_unlocked;
-	}
-
-	if (PageDirty(page))
-		ClearPageDirty(page);
-	unlock_page(page);
-
-	if (pmd_young(orig_pmd) || pmd_dirty(orig_pmd)) {
-		pmdp_invalidate(vma, addr, pmd);
-		orig_pmd = pmd_mkold(orig_pmd);
-		orig_pmd = pmd_mkclean(orig_pmd);
-
-		set_pmd_at(mm, addr, pmd, orig_pmd);
-		tlb_remove_pmd_tlb_entry(tlb, pmd, addr);
-	}
-
-	mark_page_lazyfree(page);
-	ret = true;
-out:
-	spin_unlock(ptl);
-out_unlocked:
-	return ret;
-}
-
 static inline void zap_deposited_table(struct mm_struct *mm, pmd_t *pmd)
 {
 	pgtable_t pgtable;
mm/madvise.c +135 -99
index ee210473f639..13b06dc8d402 100644
@@ -310,6 +310,91 @@ static long madvise_willneed(struct vm_area_struct *vma,
 	return 0;
 }
 
+enum madv_pmdp_reset_t {
+	MADV_PMDP_RESET,	/* pmd was reset successfully */
+	MADV_PMDP_SPLIT,	/* pmd was split */
+	MADV_PMDP_ERROR,
+};
+
+static enum madv_pmdp_reset_t madvise_pmdp_reset_or_split(struct mm_walk *walk,
+				pmd_t *pmd, spinlock_t *ptl,
+				unsigned long addr, unsigned long end,
+				bool young, bool dirty)
+{
+	pmd_t orig_pmd;
+	unsigned long next;
+	struct page *page;
+	struct mmu_gather *tlb = walk->private;
+	struct mm_struct *mm = walk->mm;
+	struct vm_area_struct *vma = walk->vma;
+	bool reset_young = false;
+	bool reset_dirty = false;
+	enum madv_pmdp_reset_t ret = MADV_PMDP_ERROR;
+
+	orig_pmd = *pmd;
+	if (is_huge_zero_pmd(orig_pmd))
+		return ret;
+
+	if (unlikely(!pmd_present(orig_pmd))) {
+		VM_BUG_ON(thp_migration_supported() &&
+				!is_pmd_migration_entry(orig_pmd));
+		return ret;
+	}
+
+	next = pmd_addr_end(addr, end);
+	page = pmd_page(orig_pmd);
+	if (next - addr != HPAGE_PMD_SIZE) {
+		/*
+		 * THP collapsing is not cheap so only split the page is
+		 * private to the this process.
+		 */
+		if (page_mapcount(page) != 1)
+			return ret;
+		get_page(page);
+		spin_unlock(ptl);
+		lock_page(page);
+		if (!split_huge_page(page))
+			ret = MADV_PMDP_SPLIT;
+		unlock_page(page);
+		put_page(page);
+		return ret;
+	}
+
+	if (young && pmd_young(orig_pmd))
+		reset_young = true;
+	if (dirty && pmd_dirty(orig_pmd))
+		reset_dirty = true;
+
+	/*
+	 * Other process could rely on the PG_dirty for data consistency,
+	 * not pte_dirty so we could reset PG_dirty only when we are owner
+	 * of the page.
+	 */
+	if (reset_dirty) {
+		if (page_mapcount(page) != 1)
+			goto out;
+		if (!trylock_page(page))
+			goto out;
+		if (PageDirty(page))
+			ClearPageDirty(page);
+		unlock_page(page);
+	}
+
+	ret = MADV_PMDP_RESET;
+	if (reset_young || reset_dirty) {
+		tlb_change_page_size(tlb, HPAGE_PMD_SIZE);
+		pmdp_invalidate(vma, addr, pmd);
+		if (reset_young)
+			orig_pmd = pmd_mkold(orig_pmd);
+		if (reset_dirty)
+			orig_pmd = pmd_mkclean(orig_pmd);
+		set_pmd_at(mm, addr, pmd, orig_pmd);
+		tlb_remove_pmd_tlb_entry(tlb, pmd, addr);
+	}
+out:
+	return ret;
+}
+
 static int madvise_cold_pte_range(pmd_t *pmd, unsigned long addr,
 				unsigned long end, struct mm_walk *walk)
 {
@@ -319,64 +404,31 @@ static int madvise_cold_pte_range(pmd_t *pmd, unsigned long addr,
 	pte_t *orig_pte, *pte, ptent;
 	spinlock_t *ptl;
 	struct page *page;
-	unsigned long next;
 
-	next = pmd_addr_end(addr, end);
 	if (pmd_trans_huge(*pmd)) {
-		pmd_t orig_pmd;
-
-		tlb_change_page_size(tlb, HPAGE_PMD_SIZE);
 		ptl = pmd_trans_huge_lock(pmd, vma);
 		if (!ptl)
 			return 0;
 
-		orig_pmd = *pmd;
-		if (is_huge_zero_pmd(orig_pmd))
-			goto huge_unlock;
-
-		if (unlikely(!pmd_present(orig_pmd))) {
-			VM_BUG_ON(thp_migration_supported() &&
-					!is_pmd_migration_entry(orig_pmd));
-			goto huge_unlock;
-		}
-
-		page = pmd_page(orig_pmd);
-		if (next - addr != HPAGE_PMD_SIZE) {
-			int err;
-
-			if (page_mapcount(page) != 1)
-				goto huge_unlock;
-
-			get_page(page);
+		switch (madvise_pmdp_reset_or_split(walk, pmd, ptl, addr, end,
+							true, false)) {
+		case MADV_PMDP_RESET:
 			spin_unlock(ptl);
-			lock_page(page);
-			err = split_huge_page(page);
-			unlock_page(page);
-			put_page(page);
-			if (!err)
-				goto regular_page;
-			return 0;
-		}
-
-		if (pmd_young(orig_pmd)) {
-			pmdp_invalidate(vma, addr, pmd);
-			orig_pmd = pmd_mkold(orig_pmd);
-
-			set_pmd_at(mm, addr, pmd, orig_pmd);
-			tlb_remove_pmd_tlb_entry(tlb, pmd, addr);
+			page = pmd_page(*pmd);
+			test_and_clear_page_young(page);
+			deactivate_page(page);
+			goto next;
+		case MADV_PMDP_ERROR:
+			spin_unlock(ptl);
+			goto next;
+		case MADV_PMDP_SPLIT:
+			; /* go through */
 		}
-
-		test_and_clear_page_young(page);
-		deactivate_page(page);
-huge_unlock:
-		spin_unlock(ptl);
-		return 0;
 	}
 
 	if (pmd_trans_unstable(pmd))
 		return 0;
 
-regular_page:
 	tlb_change_page_size(tlb, PAGE_SIZE);
 	orig_pte = pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
 	flush_tlb_batched_pending(mm);
@@ -443,6 +495,7 @@ static int madvise_cold_pte_range(pmd_t *pmd, unsigned long addr,
 
 	arch_enter_lazy_mmu_mode();
 	pte_unmap_unlock(orig_pte, ptl);
+next:
 	cond_resched();
 
 	return 0;
@@ -493,70 +546,38 @@ static int madvise_pageout_pte_range(pmd_t *pmd, unsigned long addr,
 	LIST_HEAD(page_list);
 	struct page *page;
 	int isolated = 0;
-	unsigned long next;
 
 	if (fatal_signal_pending(current))
 		return -EINTR;
 
-	next = pmd_addr_end(addr, end);
 	if (pmd_trans_huge(*pmd)) {
-		pmd_t orig_pmd;
-
-		tlb_change_page_size(tlb, HPAGE_PMD_SIZE);
 		ptl = pmd_trans_huge_lock(pmd, vma);
 		if (!ptl)
 			return 0;
 
-		orig_pmd = *pmd;
-		if (is_huge_zero_pmd(orig_pmd))
-			goto huge_unlock;
-
-		if (unlikely(!pmd_present(orig_pmd))) {
-			VM_BUG_ON(thp_migration_supported() &&
-					!is_pmd_migration_entry(orig_pmd));
-			goto huge_unlock;
-		}
-
-		page = pmd_page(orig_pmd);
-		if (next - addr != HPAGE_PMD_SIZE) {
-			int err;
-
-			if (page_mapcount(page) != 1)
-				goto huge_unlock;
-			get_page(page);
+		switch (madvise_pmdp_reset_or_split(walk, pmd, ptl, addr, end,
+							true, false)) {
+		case MADV_PMDP_RESET:
+			page = pmd_page(*pmd);
 			spin_unlock(ptl);
-			lock_page(page);
-			err = split_huge_page(page);
-			unlock_page(page);
-			put_page(page);
-			if (!err)
-				goto regular_page;
-			return 0;
-		}
-
-		if (isolate_lru_page(page))
-			goto huge_unlock;
-
-		if (pmd_young(orig_pmd)) {
-			pmdp_invalidate(vma, addr, pmd);
-			orig_pmd = pmd_mkold(orig_pmd);
-
-			set_pmd_at(mm, addr, pmd, orig_pmd);
-			tlb_remove_tlb_entry(tlb, pmd, addr);
+			if (isolate_lru_page(page))
+				return 0;
+			ClearPageReferenced(page);
+			test_and_clear_page_young(page);
+			list_add(&page->lru, &page_list);
+			reclaim_pages(&page_list);
+			goto next;
+		case MADV_PMDP_ERROR:
+			spin_unlock(ptl);
+			goto next;
+		case MADV_PMDP_SPLIT:
+			; /* go through */
 		}
-
-		ClearPageReferenced(page);
-		test_and_clear_page_young(page);
-		list_add(&page->lru, &page_list);
-huge_unlock:
-		spin_unlock(ptl);
-		reclaim_pages(&page_list);
-		return 0;
 	}
 
 	if (pmd_trans_unstable(pmd))
 		return 0;
-regular_page:
+
 	tlb_change_page_size(tlb, PAGE_SIZE);
 	orig_pte = pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
 	flush_tlb_batched_pending(mm);
@@ -631,6 +652,7 @@ static int madvise_pageout_pte_range(pmd_t *pmd, unsigned long addr,
 	arch_leave_lazy_mmu_mode();
 	pte_unmap_unlock(orig_pte, ptl);
 	reclaim_pages(&page_list);
+next:
 	cond_resched();
 
 	return 0;
@@ -700,12 +722,26 @@ static int madvise_free_pte_range(pmd_t *pmd, unsigned long addr,
 	pte_t *orig_pte, *pte, ptent;
 	struct page *page;
 	int nr_swap = 0;
-	unsigned long next;
 
-	next = pmd_addr_end(addr, end);
-	if (pmd_trans_huge(*pmd))
-		if (madvise_free_huge_pmd(tlb, vma, pmd, addr, next))
+	if (pmd_trans_huge(*pmd)) {
+		ptl = pmd_trans_huge_lock(pmd, vma);
+		if (!ptl)
+			return 0;
+
+		switch (madvise_pmdp_reset_or_split(walk, pmd, ptl, addr, end,
+							true, true)) {
+		case MADV_PMDP_RESET:
+			page = pmd_page(*pmd);
+			spin_unlock(ptl);
+			mark_page_lazyfree(page);
 			goto next;
+		case MADV_PMDP_ERROR:
+			spin_unlock(ptl);
+			goto next;
+		case MADV_PMDP_SPLIT:
+			; /* go through */
+		}
+	}
 
 	if (pmd_trans_unstable(pmd))
 		return 0;
@@ -817,8 +853,8 @@ static int madvise_free_pte_range(pmd_t *pmd, unsigned long addr,
 	}
 	arch_leave_lazy_mmu_mode();
 	pte_unmap_unlock(orig_pte, ptl);
-	cond_resched();
 next:
+	cond_resched();
 	return 0;
 }
 
Signature · 1 line
2.22.0.410.gd8fdbe21b5-goog
Signature · 2 lines
Michal Hocko
SUSE Labs
#15 lore raw

Re: [PATCH v3 4/5] mm: introduce MADV_PAGEOUT

In reply to Michal Hocko
Quoted 34 lines · Michal Hocko
Quoted 22 lines · Minchan Kim

When a process expects no accesses to a certain memory range for a long time, it could hint kernel that the pages can be reclaimed instantly but data should be preserved for future use. This could reduce workingset eviction so it ends up increasing performance.

This patch introduces the new MADV_PAGEOUT hint to madvise(2) syscall. MADV_PAGEOUT can be used by a process to mark a memory range as not expected to be used for a long time so that kernel reclaims *any LRU* pages instantly. The hint can help kernel in deciding which pages to evict proactively.

- man-page material

MADV_PAGEOUT (since Linux x.x)

Do not expect access in the near future so pages in the specified regions could be reclaimed instantly regardless of memory pressure. Thus, access in the range after successful operation could cause major page fault but never lose the up-to-date contents unlike MADV_DONTNEED.

Quoted 4 lines

It works for only private anonymous mappings and non-anonymous mappings that belong to files that the calling process could successfully open for writing; otherwise, it could be used for sidechannel attack.

I would rephrase this way:
"
Pages belonging to a shared mapping are only processed if a write access is allowed for the calling process.
"

Quoted 3 lines

I wouldn't really mention side channel attacks for a man page. You can mention can_do_mincore check and the side channel prevention in the changelog that is not aimed for the man page.

Agree. I will rephrase with one you suggested.
Thanks for the suggestion.

Quoted 25 lines
Quoted 16 lines

MADV_PAGEOUT cannot be applied to locked pages, Huge TLB pages, or VM_PFNMAP pages.

* v2
* add comment about SWAP_CLUSTER_MAX - mhocko
* add permission check to prevent sidechannel attack - mhocko
* add man page stuff - dave

* v1
* change pte to old and rely on the other's reference - hannes
* remove page_mapcount to check shared page - mhocko

* RFC v2
* make reclaim_pages simple via factoring out isolate logic - hannes

Quoted 7 lines

* RFCv1
* rename from MADV_COLD to MADV_PAGEOUT - hannes
* bail out if process is being killed - Hillf
* fix reclaim_pages bugs - Hillf

---

Quoted 6 lines

I am still not convinced about the SWAP_CLUSTER_MAX batching and the udnerlying OOM argument. Is one pmd worth of pages really an OOM risk? Sure you can have many invocations in parallel and that would add on but the same might happen with SWAP_CLUSTER_MAX. So I would just remove the batching for now and think of it only if we really see this being a problem for real. Unless you feel really strong about this, of course.

I don't have the number to support SWAP_CLUSTER_MAX batching for hinting operations. However, I wanted to be consistent with other LRU batching logic so that it could affect altogether if someone try to increase SWAP_CLUSTER_MAX which is more efficienty for batching operation, later. (AFAIK, someone tried it a few years ago but rollback soon, I couldn't rebemeber what was the reason at that time, anyway).

Quoted 4 lines

Anyway the patch looks ok to me otherwise.

Thanks!

#22 lore raw

Re: [PATCH v3 5/5] mm: factor out pmd young/dirty bit handling and THP split

In reply to Michal Hocko
Quoted 13 lines · Michal Hocko
Quoted 3 lines · Minchan Kim

Now, there are common part among MADV_COLD|PAGEOUT|FREE to reset access/dirty bit resetting or split the THP page to handle part of subpages in the THP page. This patch factor out the common part.

While this reduces the code duplication to some degree I suspect it only goes half way. I haven't tried that myself due to lack of time but I believe this has a potential to reduce even more. All those madvise calls are doing the same thing essentially. What page tables and apply an operation on ptes and/or a page that is mapped. And that suggests that the specific operation should be good with defining two - pte and page operations on each entry. All the rest should be a common code.

Quoted 3 lines

That being said, I do not feel strongly about this patch. The rest of the series should be good enough even without it and I wouldn't delay it just by discussing a potential of the cleanup.

I totally agree with you. For several cycles, some people asked me to factor common part out. I understand them why they wanted it. However, when I tried it, it's not trivial to clean it out due to subtle difference of them. If I couldn't make it clean at this moment, I want to keep them without factoing out since it's more readable, at least.

I will drop this patch next submit unless someone pop with better idea.

Quoted 431 lines
Quoted 427 lines
---
 include/linux/huge_mm.h |   3 -
 mm/huge_memory.c        |  74 -------------
 mm/madvise.c            | 234 +++++++++++++++++++++++-----------------
 3 files changed, 135 insertions(+), 176 deletions(-)
include/linux/huge_mm.h +0 -3
index 7cd5c150c21d..2667e1aa3ce5 100644
@@ -29,9 +29,6 @@ extern struct page *follow_trans_huge_pmd(struct vm_area_struct *vma,
 					  unsigned long addr,
 					  pmd_t *pmd,
 					  unsigned int flags);
-extern bool madvise_free_huge_pmd(struct mmu_gather *tlb,
-			struct vm_area_struct *vma,
-			pmd_t *pmd, unsigned long addr, unsigned long next);
 extern int zap_huge_pmd(struct mmu_gather *tlb,
 			struct vm_area_struct *vma,
 			pmd_t *pmd, unsigned long addr);
mm/huge_memory.c +0 -74
index 93f531b63a45..e4b9a06788f3 100644
@@ -1671,80 +1671,6 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf, pmd_t pmd)
 	return 0;
 }
 
-/*
- * Return true if we do MADV_FREE successfully on entire pmd page.
- * Otherwise, return false.
- */
-bool madvise_free_huge_pmd(struct mmu_gather *tlb, struct vm_area_struct *vma,
-		pmd_t *pmd, unsigned long addr, unsigned long next)
-{
-	spinlock_t *ptl;
-	pmd_t orig_pmd;
-	struct page *page;
-	struct mm_struct *mm = tlb->mm;
-	bool ret = false;
-
-	tlb_change_page_size(tlb, HPAGE_PMD_SIZE);
-
-	ptl = pmd_trans_huge_lock(pmd, vma);
-	if (!ptl)
-		goto out_unlocked;
-
-	orig_pmd = *pmd;
-	if (is_huge_zero_pmd(orig_pmd))
-		goto out;
-
-	if (unlikely(!pmd_present(orig_pmd))) {
-		VM_BUG_ON(thp_migration_supported() &&
-				  !is_pmd_migration_entry(orig_pmd));
-		goto out;
-	}
-
-	page = pmd_page(orig_pmd);
-	/*
-	 * If other processes are mapping this page, we couldn't discard
-	 * the page unless they all do MADV_FREE so let's skip the page.
-	 */
-	if (page_mapcount(page) != 1)
-		goto out;
-
-	if (!trylock_page(page))
-		goto out;
-
-	/*
-	 * If user want to discard part-pages of THP, split it so MADV_FREE
-	 * will deactivate only them.
-	 */
-	if (next - addr != HPAGE_PMD_SIZE) {
-		get_page(page);
-		spin_unlock(ptl);
-		split_huge_page(page);
-		unlock_page(page);
-		put_page(page);
-		goto out_unlocked;
-	}
-
-	if (PageDirty(page))
-		ClearPageDirty(page);
-	unlock_page(page);
-
-	if (pmd_young(orig_pmd) || pmd_dirty(orig_pmd)) {
-		pmdp_invalidate(vma, addr, pmd);
-		orig_pmd = pmd_mkold(orig_pmd);
-		orig_pmd = pmd_mkclean(orig_pmd);
-
-		set_pmd_at(mm, addr, pmd, orig_pmd);
-		tlb_remove_pmd_tlb_entry(tlb, pmd, addr);
-	}
-
-	mark_page_lazyfree(page);
-	ret = true;
-out:
-	spin_unlock(ptl);
-out_unlocked:
-	return ret;
-}
-
 static inline void zap_deposited_table(struct mm_struct *mm, pmd_t *pmd)
 {
 	pgtable_t pgtable;
mm/madvise.c +135 -99
index ee210473f639..13b06dc8d402 100644
@@ -310,6 +310,91 @@ static long madvise_willneed(struct vm_area_struct *vma,
 	return 0;
 }
 
+enum madv_pmdp_reset_t {
+	MADV_PMDP_RESET,	/* pmd was reset successfully */
+	MADV_PMDP_SPLIT,	/* pmd was split */
+	MADV_PMDP_ERROR,
+};
+
+static enum madv_pmdp_reset_t madvise_pmdp_reset_or_split(struct mm_walk *walk,
+				pmd_t *pmd, spinlock_t *ptl,
+				unsigned long addr, unsigned long end,
+				bool young, bool dirty)
+{
+	pmd_t orig_pmd;
+	unsigned long next;
+	struct page *page;
+	struct mmu_gather *tlb = walk->private;
+	struct mm_struct *mm = walk->mm;
+	struct vm_area_struct *vma = walk->vma;
+	bool reset_young = false;
+	bool reset_dirty = false;
+	enum madv_pmdp_reset_t ret = MADV_PMDP_ERROR;
+
+	orig_pmd = *pmd;
+	if (is_huge_zero_pmd(orig_pmd))
+		return ret;
+
+	if (unlikely(!pmd_present(orig_pmd))) {
+		VM_BUG_ON(thp_migration_supported() &&
+				!is_pmd_migration_entry(orig_pmd));
+		return ret;
+	}
+
+	next = pmd_addr_end(addr, end);
+	page = pmd_page(orig_pmd);
+	if (next - addr != HPAGE_PMD_SIZE) {
+		/*
+		 * THP collapsing is not cheap so only split the page is
+		 * private to the this process.
+		 */
+		if (page_mapcount(page) != 1)
+			return ret;
+		get_page(page);
+		spin_unlock(ptl);
+		lock_page(page);
+		if (!split_huge_page(page))
+			ret = MADV_PMDP_SPLIT;
+		unlock_page(page);
+		put_page(page);
+		return ret;
+	}
+
+	if (young && pmd_young(orig_pmd))
+		reset_young = true;
+	if (dirty && pmd_dirty(orig_pmd))
+		reset_dirty = true;
+
+	/*
+	 * Other process could rely on the PG_dirty for data consistency,
+	 * not pte_dirty so we could reset PG_dirty only when we are owner
+	 * of the page.
+	 */
+	if (reset_dirty) {
+		if (page_mapcount(page) != 1)
+			goto out;
+		if (!trylock_page(page))
+			goto out;
+		if (PageDirty(page))
+			ClearPageDirty(page);
+		unlock_page(page);
+	}
+
+	ret = MADV_PMDP_RESET;
+	if (reset_young || reset_dirty) {
+		tlb_change_page_size(tlb, HPAGE_PMD_SIZE);
+		pmdp_invalidate(vma, addr, pmd);
+		if (reset_young)
+			orig_pmd = pmd_mkold(orig_pmd);
+		if (reset_dirty)
+			orig_pmd = pmd_mkclean(orig_pmd);
+		set_pmd_at(mm, addr, pmd, orig_pmd);
+		tlb_remove_pmd_tlb_entry(tlb, pmd, addr);
+	}
+out:
+	return ret;
+}
+
 static int madvise_cold_pte_range(pmd_t *pmd, unsigned long addr,
 				unsigned long end, struct mm_walk *walk)
 {
@@ -319,64 +404,31 @@ static int madvise_cold_pte_range(pmd_t *pmd, unsigned long addr,
 	pte_t *orig_pte, *pte, ptent;
 	spinlock_t *ptl;
 	struct page *page;
-	unsigned long next;
 
-	next = pmd_addr_end(addr, end);
 	if (pmd_trans_huge(*pmd)) {
-		pmd_t orig_pmd;
-
-		tlb_change_page_size(tlb, HPAGE_PMD_SIZE);
 		ptl = pmd_trans_huge_lock(pmd, vma);
 		if (!ptl)
 			return 0;
 
-		orig_pmd = *pmd;
-		if (is_huge_zero_pmd(orig_pmd))
-			goto huge_unlock;
-
-		if (unlikely(!pmd_present(orig_pmd))) {
-			VM_BUG_ON(thp_migration_supported() &&
-					!is_pmd_migration_entry(orig_pmd));
-			goto huge_unlock;
-		}
-
-		page = pmd_page(orig_pmd);
-		if (next - addr != HPAGE_PMD_SIZE) {
-			int err;
-
-			if (page_mapcount(page) != 1)
-				goto huge_unlock;
-
-			get_page(page);
+		switch (madvise_pmdp_reset_or_split(walk, pmd, ptl, addr, end,
+							true, false)) {
+		case MADV_PMDP_RESET:
 			spin_unlock(ptl);
-			lock_page(page);
-			err = split_huge_page(page);
-			unlock_page(page);
-			put_page(page);
-			if (!err)
-				goto regular_page;
-			return 0;
-		}
-
-		if (pmd_young(orig_pmd)) {
-			pmdp_invalidate(vma, addr, pmd);
-			orig_pmd = pmd_mkold(orig_pmd);
-
-			set_pmd_at(mm, addr, pmd, orig_pmd);
-			tlb_remove_pmd_tlb_entry(tlb, pmd, addr);
+			page = pmd_page(*pmd);
+			test_and_clear_page_young(page);
+			deactivate_page(page);
+			goto next;
+		case MADV_PMDP_ERROR:
+			spin_unlock(ptl);
+			goto next;
+		case MADV_PMDP_SPLIT:
+			; /* go through */
 		}
-
-		test_and_clear_page_young(page);
-		deactivate_page(page);
-huge_unlock:
-		spin_unlock(ptl);
-		return 0;
 	}
 
 	if (pmd_trans_unstable(pmd))
 		return 0;
 
-regular_page:
 	tlb_change_page_size(tlb, PAGE_SIZE);
 	orig_pte = pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
 	flush_tlb_batched_pending(mm);
@@ -443,6 +495,7 @@ static int madvise_cold_pte_range(pmd_t *pmd, unsigned long addr,
 
 	arch_enter_lazy_mmu_mode();
 	pte_unmap_unlock(orig_pte, ptl);
+next:
 	cond_resched();
 
 	return 0;
@@ -493,70 +546,38 @@ static int madvise_pageout_pte_range(pmd_t *pmd, unsigned long addr,
 	LIST_HEAD(page_list);
 	struct page *page;
 	int isolated = 0;
-	unsigned long next;
 
 	if (fatal_signal_pending(current))
 		return -EINTR;
 
-	next = pmd_addr_end(addr, end);
 	if (pmd_trans_huge(*pmd)) {
-		pmd_t orig_pmd;
-
-		tlb_change_page_size(tlb, HPAGE_PMD_SIZE);
 		ptl = pmd_trans_huge_lock(pmd, vma);
 		if (!ptl)
 			return 0;
 
-		orig_pmd = *pmd;
-		if (is_huge_zero_pmd(orig_pmd))
-			goto huge_unlock;
-
-		if (unlikely(!pmd_present(orig_pmd))) {
-			VM_BUG_ON(thp_migration_supported() &&
-					!is_pmd_migration_entry(orig_pmd));
-			goto huge_unlock;
-		}
-
-		page = pmd_page(orig_pmd);
-		if (next - addr != HPAGE_PMD_SIZE) {
-			int err;
-
-			if (page_mapcount(page) != 1)
-				goto huge_unlock;
-			get_page(page);
+		switch (madvise_pmdp_reset_or_split(walk, pmd, ptl, addr, end,
+							true, false)) {
+		case MADV_PMDP_RESET:
+			page = pmd_page(*pmd);
 			spin_unlock(ptl);
-			lock_page(page);
-			err = split_huge_page(page);
-			unlock_page(page);
-			put_page(page);
-			if (!err)
-				goto regular_page;
-			return 0;
-		}
-
-		if (isolate_lru_page(page))
-			goto huge_unlock;
-
-		if (pmd_young(orig_pmd)) {
-			pmdp_invalidate(vma, addr, pmd);
-			orig_pmd = pmd_mkold(orig_pmd);
-
-			set_pmd_at(mm, addr, pmd, orig_pmd);
-			tlb_remove_tlb_entry(tlb, pmd, addr);
+			if (isolate_lru_page(page))
+				return 0;
+			ClearPageReferenced(page);
+			test_and_clear_page_young(page);
+			list_add(&page->lru, &page_list);
+			reclaim_pages(&page_list);
+			goto next;
+		case MADV_PMDP_ERROR:
+			spin_unlock(ptl);
+			goto next;
+		case MADV_PMDP_SPLIT:
+			; /* go through */
 		}
-
-		ClearPageReferenced(page);
-		test_and_clear_page_young(page);
-		list_add(&page->lru, &page_list);
-huge_unlock:
-		spin_unlock(ptl);
-		reclaim_pages(&page_list);
-		return 0;
 	}
 
 	if (pmd_trans_unstable(pmd))
 		return 0;
-regular_page:
+
 	tlb_change_page_size(tlb, PAGE_SIZE);
 	orig_pte = pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
 	flush_tlb_batched_pending(mm);
@@ -631,6 +652,7 @@ static int madvise_pageout_pte_range(pmd_t *pmd, unsigned long addr,
 	arch_leave_lazy_mmu_mode();
 	pte_unmap_unlock(orig_pte, ptl);
 	reclaim_pages(&page_list);
+next:
 	cond_resched();
 
 	return 0;
@@ -700,12 +722,26 @@ static int madvise_free_pte_range(pmd_t *pmd, unsigned long addr,
 	pte_t *orig_pte, *pte, ptent;
 	struct page *page;
 	int nr_swap = 0;
-	unsigned long next;
 
-	next = pmd_addr_end(addr, end);
-	if (pmd_trans_huge(*pmd))
-		if (madvise_free_huge_pmd(tlb, vma, pmd, addr, next))
+	if (pmd_trans_huge(*pmd)) {
+		ptl = pmd_trans_huge_lock(pmd, vma);
+		if (!ptl)
+			return 0;
+
+		switch (madvise_pmdp_reset_or_split(walk, pmd, ptl, addr, end,
+							true, true)) {
+		case MADV_PMDP_RESET:
+			page = pmd_page(*pmd);
+			spin_unlock(ptl);
+			mark_page_lazyfree(page);
 			goto next;
+		case MADV_PMDP_ERROR:
+			spin_unlock(ptl);
+			goto next;
+		case MADV_PMDP_SPLIT:
+			; /* go through */
+		}
+	}
 
 	if (pmd_trans_unstable(pmd))
 		return 0;
@@ -817,8 +853,8 @@ static int madvise_free_pte_range(pmd_t *pmd, unsigned long addr,
 	}
 	arch_leave_lazy_mmu_mode();
 	pte_unmap_unlock(orig_pte, ptl);
-	cond_resched();
 next:
+	cond_resched();
 	return 0;
 }
 
Signature · 1 line
2.22.0.410.gd8fdbe21b5-goog
Signature · 2 lines
Michal Hocko
SUSE Labs
reviewer
#16 lore raw

Re: [PATCH v3 4/5] mm: introduce MADV_PAGEOUT

In reply to Minchan Kim
Quoted 1 line · Minchan Kim

On Tue, Jul 09, 2019 at 11:55:19AM +0200, Michal Hocko wrote:

[...]

Quoted 7 lines
Quoted 6 lines

I am still not convinced about the SWAP_CLUSTER_MAX batching and the udnerlying OOM argument. Is one pmd worth of pages really an OOM risk? Sure you can have many invocations in parallel and that would add on but the same might happen with SWAP_CLUSTER_MAX. So I would just remove the batching for now and think of it only if we really see this being a problem for real. Unless you feel really strong about this, of course.

Quoted 6 lines

I don't have the number to support SWAP_CLUSTER_MAX batching for hinting operations. However, I wanted to be consistent with other LRU batching logic so that it could affect altogether if someone try to increase SWAP_CLUSTER_MAX which is more efficienty for batching operation, later. (AFAIK, someone tried it a few years ago but rollback soon, I couldn't rebemeber what was the reason at that time, anyway).

Then please drop this part. It makes the code more complex while any benefit is not demonstrated.

Quoted 5 lines
Quoted 3 lines

Anyway the patch looks ok to me otherwise.

Thanks!

Signature · 2 lines
Michal Hocko
SUSE Labs
#17 lore raw

Re: [PATCH v3 4/5] mm: introduce MADV_PAGEOUT

Quoted 17 lines · Michal Hocko
Quoted 1 line · Minchan Kim

On Tue, Jul 09, 2019 at 11:55:19AM +0200, Michal Hocko wrote:

[...]

Quoted 7 lines
Quoted 6 lines

I am still not convinced about the SWAP_CLUSTER_MAX batching and the udnerlying OOM argument. Is one pmd worth of pages really an OOM risk? Sure you can have many invocations in parallel and that would add on but the same might happen with SWAP_CLUSTER_MAX. So I would just remove the batching for now and think of it only if we really see this being a problem for real. Unless you feel really strong about this, of course.

Quoted 6 lines

I don't have the number to support SWAP_CLUSTER_MAX batching for hinting operations. However, I wanted to be consistent with other LRU batching logic so that it could affect altogether if someone try to increase SWAP_CLUSTER_MAX which is more efficienty for batching operation, later. (AFAIK, someone tried it a few years ago but rollback soon, I couldn't rebemeber what was the reason at that time, anyway).

Quoted 2 lines

Then please drop this part. It makes the code more complex while any benefit is not demonstrated.

The history says the benefit.
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/patch/?id=d37dd5dcb955dd8c2cdd4eaef1f15d1b7ecbc379
With the history, rather than proving it's worth for upcoming new code, need to try to prove no harmful any longer if we want to remove(or not consistent with other reclaim path). It's not the goal of this patch.

reviewer
#18 lore raw

Re: [PATCH v3 4/5] mm: introduce MADV_PAGEOUT

Quoted 18 lines · Minchan Kim
Quoted 17 lines · Michal Hocko
Quoted 1 line · Minchan Kim

On Tue, Jul 09, 2019 at 11:55:19AM +0200, Michal Hocko wrote:

[...]

Quoted 13 lines
Quoted 6 lines

I am still not convinced about the SWAP_CLUSTER_MAX batching and the udnerlying OOM argument. Is one pmd worth of pages really an OOM risk? Sure you can have many invocations in parallel and that would add on but the same might happen with SWAP_CLUSTER_MAX. So I would just remove the batching for now and think of it only if we really see this being a problem for real. Unless you feel really strong about this, of course.

I don't have the number to support SWAP_CLUSTER_MAX batching for hinting operations. However, I wanted to be consistent with other LRU batching logic so that it could affect altogether if someone try to increase SWAP_CLUSTER_MAX which is more efficienty for batching operation, later. (AFAIK, someone tried it a few years ago but rollback soon, I couldn't rebemeber what was the reason at that time, anyway).

Quoted 5 lines
Quoted 2 lines

Then please drop this part. It makes the code more complex while any benefit is not demonstrated.

The history says the benefit.
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/patch/?id=d37dd5dcb955dd8c2cdd4eaef1f15d1b7ecbc379

Limiting the number of isolated pages is fine. All I am saying is that SWAP_CLUSTER_MAX is an arbitrary number same as 512 pages for one PMD as a unit of work. Both can lead to the same effect if there are too many parallel tasks doing the same thing.

I do not want you to change that in the reclaim path. All I am asking for is to add a bathing without any actual data to back that because that makes the code more complex without any gains.

Signature · 2 lines
Michal Hocko
SUSE Labs
#19 lore raw

Re: [PATCH v3 4/5] mm: introduce MADV_PAGEOUT

Quoted 25 lines · Michal Hocko
Quoted 18 lines · Minchan Kim
Quoted 17 lines · Michal Hocko
Quoted 1 line · Minchan Kim

On Tue, Jul 09, 2019 at 11:55:19AM +0200, Michal Hocko wrote:

[...]

Quoted 13 lines
Quoted 6 lines

I am still not convinced about the SWAP_CLUSTER_MAX batching and the udnerlying OOM argument. Is one pmd worth of pages really an OOM risk? Sure you can have many invocations in parallel and that would add on but the same might happen with SWAP_CLUSTER_MAX. So I would just remove the batching for now and think of it only if we really see this being a problem for real. Unless you feel really strong about this, of course.

I don't have the number to support SWAP_CLUSTER_MAX batching for hinting operations. However, I wanted to be consistent with other LRU batching logic so that it could affect altogether if someone try to increase SWAP_CLUSTER_MAX which is more efficienty for batching operation, later. (AFAIK, someone tried it a few years ago but rollback soon, I couldn't rebemeber what was the reason at that time, anyway).

Quoted 5 lines
Quoted 2 lines

Then please drop this part. It makes the code more complex while any benefit is not demonstrated.

The history says the benefit.
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/patch/?id=d37dd5dcb955dd8c2cdd4eaef1f15d1b7ecbc379

Quoted 8 lines

Limiting the number of isolated pages is fine. All I am saying is that SWAP_CLUSTER_MAX is an arbitrary number same as 512 pages for one PMD as a unit of work. Both can lead to the same effect if there are too many parallel tasks doing the same thing.

I do not want you to change that in the reclaim path. All I am asking for is to add a bathing without any actual data to back that because that makes the code more complex without any gains.

I understand what you meant and I'm really one to make code simple. However, my concern was that we have isolated by SWAP_CLUSTER_MAX(32 pages) for other path(reclaim/compaction) so I want to be consistent with others. If you think that the consistency(IOW, others are 32 limit but here 256 limit) is no helpful this case, I don't have any strong opinion. Let's drop the part. I will add it into description, then.

Thanks.

Quoted 3 lines
Signature · 2 lines
Michal Hocko
SUSE Labs