[PATCH v2 00/12] mm: userspace hugepage collapse

[PATCH v2 00/12] mm: userspace hugepage collapse

[Zach O'Keefe]

Introduction
--------------------------------

This series provides a mechanism for userspace to induce a collapse of
eligible ranges of memory into transparent hugepages in process context,
thus permitting users to more tightly control their own hugepage
utilization policy at their own expense.

This idea was introduced by David Rientjes[1], and the semantics and
implementation were introduced and discussed in a previous PATCH RFC[2].

Interface
--------------------------------

The proposed interface adds a new madvise(2) mode, MADV_COLLAPSE, and
leverages the new process_madvise(2) call.

process_madvise(2)

	Performs a synchronous collapse of the native pages
	mapped by the list of iovecs into transparent hugepages.

	Allocation semantics are the same as khugepaged, and depend on
	(1) the active sysfs settings
	/sys/kernel/mm/transparent_hugepage/enabled and
	/sys/kernel/mm/transparent_hugepage/khugepaged/defrag, and (2)
	the VMA flags of the memory range being collapsed.

	Collapse eligibility criteria differs from khugepaged in that
	the sysfs files
	/sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_[none|swap|shared]
	are ignored.

	When a range spans multiple VMAs, the semantics of the collapse over
	of each VMA is independent from the others.

	Caller must have CAP_SYS_ADMIN if not acting on self.

	Return value follows existing process_madvise(2) conventions.  A
	“success” indicates that all hugepage-sized/aligned regions
	covered by the provided range were either successfully
	collapsed, or were already pmd-mapped THPs.

madvise(2)

	Equivalent to process_madvise(2) on self, with 0 returned on
	“success”.

Future work
--------------------------------

Only private anonymous memory is supported by this series. File and
shmem memory support will be added later.

One possible user of this functionality is a userspace agent that
attempts to optimize THP utilization system-wide by allocating THPs
based on, for example, task priority, task performance requirements, or
heatmaps.  For the latter, one idea that has already surfaced is using
DAMON to identify hot regions, and driving THP collapse through a new
DAMOS_COLLAPSE scheme[3].

Sequence of Patches
--------------------------------

Patches 1-4 perform refactoring of collapse logic within khugepaged.c
and introduce the notion of a collapse context.

Patches 5-9 introduces MADV_COLLAPSE, does some renaming, adds support
so that MADV_COLLAPSE context has the eligibility and allocation
semantics referenced above, and adds process_madivse(2) support.

Patches 10-12 add selftests to test the new functionality.

Applies against next-20220414.

v1 -> v2:
* Cover-letter clarification and added RFC -> v1 notes
* Fixes issues reported by kernel test robot <lkp@intel.com>
* 'mm/khugepaged: record SCAN_PMD_MAPPED when scan_pmd() finds THP'
  -> Fixed mixed code/declarations
* 'mm/khugepaged: make hugepage allocation context-specific'
  -> Fixed bad function signature in !NUMA && TRANSPARENT_HUGEPAGE configs
  -> Added doc comment to retract_page_tables() for "cc"
* 'mm/khugepaged: add struct collapse_result'
  -> Added doc comment to retract_page_tables() for "cr"
* 'mm/madvise: introduce MADV_COLLAPSE sync hugepage collapse'
  -> Added MADV_COLLAPSE definitions for alpha, mips, parisc, xtensa
  -> Moved an "#ifdef NUMA" so that khugepaged_find_target_node() is
     defined in !NUMA && TRANSPARENT_HUGEPAGE configs.
* 'mm/khugepaged: remove khugepaged prefix from shared collapse'
  functions
  -> Removed khugepaged prefix from khugepaged_find_target_node on L914
* Rebased onto next-20220414

RFC -> v1:
* The series was significantly reworked from RFC and most patches are entirely
  new or reworked.
* Collapse eligibility criteria has changed: MADV_COLLAPSE now respects
  VM_NOHUGEPAGE.
* Collapse semantics have changed: the gfp flags used for hugepage allocation
  now match that of khugepaged for the same VMA, instead of the gfp flags used
  at-fault for calling process for the VMA.
* Collapse semantics have changed: The collapse semantics for multiple VMAs
  spanning a single MADV_COLLAPSE call are now independent, whereas before the
  idea was to allow direct reclaim/compaction if any spanned VMA permitted so.
* The process_madvise(2) flags, MADV_F_COLLAPSE_LIMITS and
  MADV_F_COLLAPSE_DEFRAG have been removed.
* Implementation change: the RFC implemented collapse over a range of hugepages
  in a batched-fashion with the aim of doing multiple page table updates inside
  a single mmap_lock write.  This has been changed, and the implementation now
  collapses each hugepage-aligned/sized region iteratively.  This was motivated
  by an experiment which showed that, when multiple threads were concurrently
  faulting during a MADV_COLLAPSE operation, mean and tail latency to acquire
  mmap_lock in read for threads in the fault patch was improved by using a batch
  size of 1 (batch sizes of 1, 8, 16, 32 were tested)[4].
* Added: If a collapse operation fails because a page isn't found on the LRU, do
  a lru_add_drain_all() and retry.
* Added: selftests

[1] https://lore.kernel.org/all/C8C89F13-3F04-456B-BA76-DE2C378D30BF@nvidia.com/
[2] https://lore.kernel.org/linux-mm/20220308213417.1407042-1-zokeefe@google.com/
[3] https://lore.kernel.org/lkml/bcc8d9a0-81d-5f34-5e4-fcc28eb7ce@google.com/T/
[4] https://lore.kernel.org/linux-mm/CAAa6QmRc76n-dspGT7UK8DkaqZAOz-CkCsME1V7KGtQ6Yt2FqA@mail.gmail.com/

Zach O'Keefe (13):
  mm/khugepaged: separate hugepage preallocation and cleanup
  mm/khugepaged: record SCAN_PMD_MAPPED when scan_pmd() finds THP
  mm/khugepaged: add struct collapse_control
  mm/khugepaged: make hugepage allocation context-specific
  mm/khugepaged: add struct collapse_result
  mm/madvise: introduce MADV_COLLAPSE sync hugepage collapse
  mm/khugepaged: remove khugepaged prefix from shared collapse functions
  mm/khugepaged: add flag to ignore khugepaged_max_ptes_*
  mm/khugepaged: add flag to ignore page young/referenced requirement
  mm/madvise: add MADV_COLLAPSE to process_madvise()
  selftests/vm: modularize collapse selftests
  selftests/vm: add MADV_COLLAPSE collapse context to selftests
  selftests/vm: add test to verify recollapse of THPs

 include/linux/huge_mm.h                 |  12 +
 include/trace/events/huge_memory.h      |   5 +-
 include/uapi/asm-generic/mman-common.h  |   2 +
 mm/internal.h                           |   1 +
 mm/khugepaged.c                         | 598 ++++++++++++++++--------
 mm/madvise.c                            |  11 +-
 mm/rmap.c                               |  15 +-
 tools/testing/selftests/vm/khugepaged.c | 417 +++++++++++------
 8 files changed, 702 insertions(+), 359 deletions(-)

-- 
2.35.1.1178.g4f1659d476-goog

[PATCH v2 01/12] mm/khugepaged: record SCAN_PMD_MAPPED when scan_pmd() finds THP

[Zach O'Keefe]

When scanning an anon pmd to see if it's eligible for collapse, return
SCAN_PMD_MAPPED if the pmd already maps a THP.  Note that
SCAN_PMD_MAPPED is different from SCAN_PAGE_COMPOUND used in the
file-collapse path, since the latter might identify pte-mapped compound
pages.  This is required by MADV_COLLAPSE which necessarily needs to
know what hugepage-aligned/sized regions are already pmd-mapped.

Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reported-by: kernel test robot <lkp@intel.com>
---
 include/trace/events/huge_memory.h |  3 ++-
 mm/internal.h                      |  1 +
 mm/khugepaged.c                    | 30 ++++++++++++++++++++++++++----
 mm/rmap.c                          | 15 +++++++++++++--
 4 files changed, 42 insertions(+), 7 deletions(-)

diff --git a/include/trace/events/huge_memory.h b/include/trace/events/huge_memory.h
index d651f3437367..9faa678e0a5b 100644
--- a/include/trace/events/huge_memory.h
+++ b/include/trace/events/huge_memory.h
@@ -33,7 +33,8 @@
 	EM( SCAN_ALLOC_HUGE_PAGE_FAIL,	"alloc_huge_page_failed")	\
 	EM( SCAN_CGROUP_CHARGE_FAIL,	"ccgroup_charge_failed")	\
 	EM( SCAN_TRUNCATED,		"truncated")			\
-	EMe(SCAN_PAGE_HAS_PRIVATE,	"page_has_private")		\
+	EM( SCAN_PAGE_HAS_PRIVATE,	"page_has_private")		\
+	EMe(SCAN_PMD_MAPPED,		"page_pmd_mapped")		\
 
 #undef EM
 #undef EMe
diff --git a/mm/internal.h b/mm/internal.h
index 48eb2d24fcd2..24fca92bd51a 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -173,6 +173,7 @@ extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason
 /*
  * in mm/rmap.c:
  */
+pmd_t *mm_find_pmd_raw(struct mm_struct *mm, unsigned long address);
 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
 
 /*
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index cb43c3aee8b2..5e5404aa6579 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -51,6 +51,7 @@ enum scan_result {
 	SCAN_CGROUP_CHARGE_FAIL,
 	SCAN_TRUNCATED,
 	SCAN_PAGE_HAS_PRIVATE,
+	SCAN_PMD_MAPPED,
 };
 
 #define CREATE_TRACE_POINTS
@@ -987,6 +988,29 @@ static int hugepage_vma_revalidate(struct mm_struct *mm, unsigned long address,
 	return 0;
 }
 
+static int find_pmd_or_thp_or_none(struct mm_struct *mm,
+				   unsigned long address,
+				   pmd_t **pmd)
+{
+	pmd_t pmde;
+
+	*pmd = mm_find_pmd_raw(mm, address);
+	if (!*pmd)
+		return SCAN_PMD_NULL;
+
+	pmde = pmd_read_atomic(*pmd);
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+	/* See comments in pmd_none_or_trans_huge_or_clear_bad() */
+	barrier();
+#endif
+	if (!pmd_present(pmde) || pmd_none(pmde))
+		return SCAN_PMD_NULL;
+	if (pmd_trans_huge(pmde))
+		return SCAN_PMD_MAPPED;
+	return SCAN_SUCCEED;
+}
+
 /*
  * Bring missing pages in from swap, to complete THP collapse.
  * Only done if khugepaged_scan_pmd believes it is worthwhile.
@@ -1238,11 +1262,9 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 
 	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
 
-	pmd = mm_find_pmd(mm, address);
-	if (!pmd) {
-		result = SCAN_PMD_NULL;
+	result = find_pmd_or_thp_or_none(mm, address, &pmd);
+	if (result != SCAN_SUCCEED)
 		goto out;
-	}
 
 	memset(khugepaged_node_load, 0, sizeof(khugepaged_node_load));
 	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
diff --git a/mm/rmap.c b/mm/rmap.c
index edfe61f95a7f..bf2a3a08d965 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -759,13 +759,12 @@ unsigned long page_address_in_vma(struct page *page, struct vm_area_struct *vma)
 	return vma_address(page, vma);
 }
 
-pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address)
+pmd_t *mm_find_pmd_raw(struct mm_struct *mm, unsigned long address)
 {
 	pgd_t *pgd;
 	p4d_t *p4d;
 	pud_t *pud;
 	pmd_t *pmd = NULL;
-	pmd_t pmde;
 
 	pgd = pgd_offset(mm, address);
 	if (!pgd_present(*pgd))
@@ -780,6 +779,18 @@ pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address)
 		goto out;
 
 	pmd = pmd_offset(pud, address);
+out:
+	return pmd;
+}
+
+pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address)
+{
+	pmd_t pmde;
+	pmd_t *pmd;
+
+	pmd = mm_find_pmd_raw(mm, address);
+	if (!pmd)
+		goto out;
 	/*
 	 * Some THP functions use the sequence pmdp_huge_clear_flush(), set_pmd_at()
 	 * without holding anon_vma lock for write.  So when looking for a
-- 
2.36.0.rc0.470.gd361397f0d-goog

[PATCH v2 02/12] mm/khugepaged: add struct collapse_control

[Zach O'Keefe]

Modularize hugepage collapse by introducing struct collapse_control.
This structure serves to describe the properties of the requested
collapse, as well as serve as a local scratch pad to use during the
collapse itself.

Signed-off-by: Zach O'Keefe <zokeefe@google.com>
---
 mm/khugepaged.c | 79 ++++++++++++++++++++++++++++---------------------
 1 file changed, 46 insertions(+), 33 deletions(-)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 5e5404aa6579..25f45ac7f6bd 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -86,6 +86,14 @@ static struct kmem_cache *mm_slot_cache __read_mostly;
 
 #define MAX_PTE_MAPPED_THP 8
 
+struct collapse_control {
+	/* Num pages scanned per node */
+	int node_load[MAX_NUMNODES];
+
+	/* Last target selected in khugepaged_find_target_node() for this scan */
+	int last_target_node;
+};
+
 /**
  * struct mm_slot - hash lookup from mm to mm_slot
  * @hash: hash collision list
@@ -796,9 +804,7 @@ static void khugepaged_alloc_sleep(void)
 	remove_wait_queue(&khugepaged_wait, &wait);
 }
 
-static int khugepaged_node_load[MAX_NUMNODES];
-
-static bool khugepaged_scan_abort(int nid)
+static bool khugepaged_scan_abort(int nid, struct collapse_control *cc)
 {
 	int i;
 
@@ -810,11 +816,11 @@ static bool khugepaged_scan_abort(int nid)
 		return false;
 
 	/* If there is a count for this node already, it must be acceptable */
-	if (khugepaged_node_load[nid])
+	if (cc->node_load[nid])
 		return false;
 
 	for (i = 0; i < MAX_NUMNODES; i++) {
-		if (!khugepaged_node_load[i])
+		if (!cc->node_load[i])
 			continue;
 		if (node_distance(nid, i) > node_reclaim_distance)
 			return true;
@@ -829,28 +835,28 @@ static inline gfp_t alloc_hugepage_khugepaged_gfpmask(void)
 }
 
 #ifdef CONFIG_NUMA
-static int khugepaged_find_target_node(void)
+static int khugepaged_find_target_node(struct collapse_control *cc)
 {
-	static int last_khugepaged_target_node = NUMA_NO_NODE;
 	int nid, target_node = 0, max_value = 0;
 
 	/* find first node with max normal pages hit */
 	for (nid = 0; nid < MAX_NUMNODES; nid++)
-		if (khugepaged_node_load[nid] > max_value) {
-			max_value = khugepaged_node_load[nid];
+		if (cc->node_load[nid] > max_value) {
+			max_value = cc->node_load[nid];
 			target_node = nid;
 		}
 
 	/* do some balance if several nodes have the same hit record */
-	if (target_node <= last_khugepaged_target_node)
-		for (nid = last_khugepaged_target_node + 1; nid < MAX_NUMNODES;
-				nid++)
-			if (max_value == khugepaged_node_load[nid]) {
+	if (target_node <= cc->last_target_node)
+		for (nid = cc->last_target_node + 1; nid < MAX_NUMNODES;
+		     nid++) {
+			if (max_value == cc->node_load[nid]) {
 				target_node = nid;
 				break;
 			}
+		}
 
-	last_khugepaged_target_node = target_node;
+	cc->last_target_node = target_node;
 	return target_node;
 }
 
@@ -888,7 +894,7 @@ khugepaged_alloc_page(struct page **hpage, gfp_t gfp, int node)
 	return *hpage;
 }
 #else
-static int khugepaged_find_target_node(void)
+static int khugepaged_find_target_node(struct collapse_control *cc)
 {
 	return 0;
 }
@@ -1248,7 +1254,8 @@ static void collapse_huge_page(struct mm_struct *mm,
 static int khugepaged_scan_pmd(struct mm_struct *mm,
 			       struct vm_area_struct *vma,
 			       unsigned long address,
-			       struct page **hpage)
+			       struct page **hpage,
+			       struct collapse_control *cc)
 {
 	pmd_t *pmd;
 	pte_t *pte, *_pte;
@@ -1266,7 +1273,7 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 	if (result != SCAN_SUCCEED)
 		goto out;
 
-	memset(khugepaged_node_load, 0, sizeof(khugepaged_node_load));
+	memset(cc->node_load, 0, sizeof(cc->node_load));
 	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
 	for (_address = address, _pte = pte; _pte < pte+HPAGE_PMD_NR;
 	     _pte++, _address += PAGE_SIZE) {
@@ -1332,16 +1339,16 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 
 		/*
 		 * Record which node the original page is from and save this
-		 * information to khugepaged_node_load[].
+		 * information to cc->node_load[].
 		 * Khugepaged will allocate hugepage from the node has the max
 		 * hit record.
 		 */
 		node = page_to_nid(page);
-		if (khugepaged_scan_abort(node)) {
+		if (khugepaged_scan_abort(node, cc)) {
 			result = SCAN_SCAN_ABORT;
 			goto out_unmap;
 		}
-		khugepaged_node_load[node]++;
+		cc->node_load[node]++;
 		if (!PageLRU(page)) {
 			result = SCAN_PAGE_LRU;
 			goto out_unmap;
@@ -1392,7 +1399,7 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 out_unmap:
 	pte_unmap_unlock(pte, ptl);
 	if (ret) {
-		node = khugepaged_find_target_node();
+		node = khugepaged_find_target_node(cc);
 		/* collapse_huge_page will return with the mmap_lock released */
 		collapse_huge_page(mm, address, hpage, node,
 				referenced, unmapped);
@@ -2044,7 +2051,8 @@ static void collapse_file(struct mm_struct *mm,
 }
 
 static void khugepaged_scan_file(struct mm_struct *mm,
-		struct file *file, pgoff_t start, struct page **hpage)
+		struct file *file, pgoff_t start, struct page **hpage,
+		struct collapse_control *cc)
 {
 	struct page *page = NULL;
 	struct address_space *mapping = file->f_mapping;
@@ -2055,7 +2063,7 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 
 	present = 0;
 	swap = 0;
-	memset(khugepaged_node_load, 0, sizeof(khugepaged_node_load));
+	memset(cc->node_load, 0, sizeof(cc->node_load));
 	rcu_read_lock();
 	xas_for_each(&xas, page, start + HPAGE_PMD_NR - 1) {
 		if (xas_retry(&xas, page))
@@ -2080,11 +2088,11 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 		}
 
 		node = page_to_nid(page);
-		if (khugepaged_scan_abort(node)) {
+		if (khugepaged_scan_abort(node, cc)) {
 			result = SCAN_SCAN_ABORT;
 			break;
 		}
-		khugepaged_node_load[node]++;
+		cc->node_load[node]++;
 
 		if (!PageLRU(page)) {
 			result = SCAN_PAGE_LRU;
@@ -2117,7 +2125,7 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 			result = SCAN_EXCEED_NONE_PTE;
 			count_vm_event(THP_SCAN_EXCEED_NONE_PTE);
 		} else {
-			node = khugepaged_find_target_node();
+			node = khugepaged_find_target_node(cc);
 			collapse_file(mm, file, start, hpage, node);
 		}
 	}
@@ -2126,7 +2134,8 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 }
 #else
 static void khugepaged_scan_file(struct mm_struct *mm,
-		struct file *file, pgoff_t start, struct page **hpage)
+		struct file *file, pgoff_t start, struct page **hpage,
+		struct collapse_control *cc)
 {
 	BUILD_BUG();
 }
@@ -2137,7 +2146,8 @@ static void khugepaged_collapse_pte_mapped_thps(struct mm_slot *mm_slot)
 #endif
 
 static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
-					    struct page **hpage)
+					    struct page **hpage,
+					    struct collapse_control *cc)
 	__releases(&khugepaged_mm_lock)
 	__acquires(&khugepaged_mm_lock)
 {
@@ -2213,12 +2223,12 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
 
 				mmap_read_unlock(mm);
 				ret = 1;
-				khugepaged_scan_file(mm, file, pgoff, hpage);
+				khugepaged_scan_file(mm, file, pgoff, hpage, cc);
 				fput(file);
 			} else {
 				ret = khugepaged_scan_pmd(mm, vma,
 						khugepaged_scan.address,
-						hpage);
+						hpage, cc);
 			}
 			/* move to next address */
 			khugepaged_scan.address += HPAGE_PMD_SIZE;
@@ -2274,7 +2284,7 @@ static int khugepaged_wait_event(void)
 		kthread_should_stop();
 }
 
-static void khugepaged_do_scan(void)
+static void khugepaged_do_scan(struct collapse_control *cc)
 {
 	struct page *hpage = NULL;
 	unsigned int progress = 0, pass_through_head = 0;
@@ -2298,7 +2308,7 @@ static void khugepaged_do_scan(void)
 		if (khugepaged_has_work() &&
 		    pass_through_head < 2)
 			progress += khugepaged_scan_mm_slot(pages - progress,
-							    &hpage);
+							    &hpage, cc);
 		else
 			progress = pages;
 		spin_unlock(&khugepaged_mm_lock);
@@ -2337,12 +2347,15 @@ static void khugepaged_wait_work(void)
 static int khugepaged(void *none)
 {
 	struct mm_slot *mm_slot;
+	struct collapse_control cc = {
+		.last_target_node = NUMA_NO_NODE,
+	};
 
 	set_freezable();
 	set_user_nice(current, MAX_NICE);
 
 	while (!kthread_should_stop()) {
-		khugepaged_do_scan();
+		khugepaged_do_scan(&cc);
 		khugepaged_wait_work();
 	}
 
-- 
2.36.0.rc0.470.gd361397f0d-goog

[PATCH v2 03/12] mm/khugepaged: make hugepage allocation context-specific

[Zach O'Keefe]

Add hugepage allocation context to struct collapse_context, allowing
different collapse contexts to allocate hugepages differently.  For
example, khugepaged decides to allocate differently in NUMA and UMA
configurations, and other collapse contexts shouldn't be coupled to this
decision.

Additionally, move [pre]allocated hugepage pointer into
struct collapse_context.

Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reported-by: kernel test robot <lkp@intel.com>
---
 mm/khugepaged.c | 96 ++++++++++++++++++++++++-------------------------
 1 file changed, 48 insertions(+), 48 deletions(-)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 25f45ac7f6bd..21c8436fa73c 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -92,6 +92,10 @@ struct collapse_control {
 
 	/* Last target selected in khugepaged_find_target_node() for this scan */
 	int last_target_node;
+
+	struct page *hpage;
+	struct page* (*alloc_hpage)(struct collapse_control *cc, gfp_t gfp,
+				    int node);
 };
 
 /**
@@ -877,21 +881,21 @@ static bool khugepaged_prealloc_page(struct page **hpage, bool *wait)
 	return true;
 }
 
-static struct page *
-khugepaged_alloc_page(struct page **hpage, gfp_t gfp, int node)
+static struct page *khugepaged_alloc_page(struct collapse_control *cc,
+					  gfp_t gfp, int node)
 {
-	VM_BUG_ON_PAGE(*hpage, *hpage);
+	VM_BUG_ON_PAGE(cc->hpage, cc->hpage);
 
-	*hpage = __alloc_pages_node(node, gfp, HPAGE_PMD_ORDER);
-	if (unlikely(!*hpage)) {
+	cc->hpage = __alloc_pages_node(node, gfp, HPAGE_PMD_ORDER);
+	if (unlikely(!cc->hpage)) {
 		count_vm_event(THP_COLLAPSE_ALLOC_FAILED);
-		*hpage = ERR_PTR(-ENOMEM);
+		cc->hpage = ERR_PTR(-ENOMEM);
 		return NULL;
 	}
 
-	prep_transhuge_page(*hpage);
+	prep_transhuge_page(cc->hpage);
 	count_vm_event(THP_COLLAPSE_ALLOC);
-	return *hpage;
+	return cc->hpage;
 }
 #else
 static int khugepaged_find_target_node(struct collapse_control *cc)
@@ -953,12 +957,12 @@ static bool khugepaged_prealloc_page(struct page **hpage, bool *wait)
 	return true;
 }
 
-static struct page *
-khugepaged_alloc_page(struct page **hpage, gfp_t gfp, int node)
+static struct page *khugepaged_alloc_page(struct collapse_control *cc,
+					  gfp_t gfp, int node)
 {
-	VM_BUG_ON(!*hpage);
+	VM_BUG_ON(!cc->hpage);
 
-	return  *hpage;
+	return cc->hpage;
 }
 #endif
 
@@ -1080,10 +1084,9 @@ static bool __collapse_huge_page_swapin(struct mm_struct *mm,
 	return true;
 }
 
-static void collapse_huge_page(struct mm_struct *mm,
-				   unsigned long address,
-				   struct page **hpage,
-				   int node, int referenced, int unmapped)
+static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
+			       struct collapse_control *cc, int referenced,
+			       int unmapped)
 {
 	LIST_HEAD(compound_pagelist);
 	pmd_t *pmd, _pmd;
@@ -1096,6 +1099,7 @@ static void collapse_huge_page(struct mm_struct *mm,
 	struct mmu_notifier_range range;
 	gfp_t gfp;
 	const struct cpumask *cpumask;
+	int node;
 
 	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
 
@@ -1110,13 +1114,14 @@ static void collapse_huge_page(struct mm_struct *mm,
 	 */
 	mmap_read_unlock(mm);
 
+	node = khugepaged_find_target_node(cc);
 	/* sched to specified node before huage page memory copy */
 	if (task_node(current) != node) {
 		cpumask = cpumask_of_node(node);
 		if (!cpumask_empty(cpumask))
 			set_cpus_allowed_ptr(current, cpumask);
 	}
-	new_page = khugepaged_alloc_page(hpage, gfp, node);
+	new_page = cc->alloc_hpage(cc, gfp, node);
 	if (!new_page) {
 		result = SCAN_ALLOC_HUGE_PAGE_FAIL;
 		goto out_nolock;
@@ -1238,15 +1243,15 @@ static void collapse_huge_page(struct mm_struct *mm,
 	update_mmu_cache_pmd(vma, address, pmd);
 	spin_unlock(pmd_ptl);
 
-	*hpage = NULL;
+	cc->hpage = NULL;
 
 	khugepaged_pages_collapsed++;
 	result = SCAN_SUCCEED;
 out_up_write:
 	mmap_write_unlock(mm);
 out_nolock:
-	if (!IS_ERR_OR_NULL(*hpage))
-		mem_cgroup_uncharge(page_folio(*hpage));
+	if (!IS_ERR_OR_NULL(cc->hpage))
+		mem_cgroup_uncharge(page_folio(cc->hpage));
 	trace_mm_collapse_huge_page(mm, isolated, result);
 	return;
 }
@@ -1254,7 +1259,6 @@ static void collapse_huge_page(struct mm_struct *mm,
 static int khugepaged_scan_pmd(struct mm_struct *mm,
 			       struct vm_area_struct *vma,
 			       unsigned long address,
-			       struct page **hpage,
 			       struct collapse_control *cc)
 {
 	pmd_t *pmd;
@@ -1399,10 +1403,8 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 out_unmap:
 	pte_unmap_unlock(pte, ptl);
 	if (ret) {
-		node = khugepaged_find_target_node(cc);
 		/* collapse_huge_page will return with the mmap_lock released */
-		collapse_huge_page(mm, address, hpage, node,
-				referenced, unmapped);
+		collapse_huge_page(mm, address, cc, referenced, unmapped);
 	}
 out:
 	trace_mm_khugepaged_scan_pmd(mm, page, writable, referenced,
@@ -1667,8 +1669,7 @@ static void retract_page_tables(struct address_space *mapping, pgoff_t pgoff)
  * @mm: process address space where collapse happens
  * @file: file that collapse on
  * @start: collapse start address
- * @hpage: new allocated huge page for collapse
- * @node: appointed node the new huge page allocate from
+ * @cc: collapse context and scratchpad
  *
  * Basic scheme is simple, details are more complex:
  *  - allocate and lock a new huge page;
@@ -1686,8 +1687,8 @@ static void retract_page_tables(struct address_space *mapping, pgoff_t pgoff)
  *    + unlock and free huge page;
  */
 static void collapse_file(struct mm_struct *mm,
-		struct file *file, pgoff_t start,
-		struct page **hpage, int node)
+			  struct file *file, pgoff_t start,
+			  struct collapse_control *cc)
 {
 	struct address_space *mapping = file->f_mapping;
 	gfp_t gfp;
@@ -1697,15 +1698,16 @@ static void collapse_file(struct mm_struct *mm,
 	XA_STATE_ORDER(xas, &mapping->i_pages, start, HPAGE_PMD_ORDER);
 	int nr_none = 0, result = SCAN_SUCCEED;
 	bool is_shmem = shmem_file(file);
-	int nr;
+	int nr, node;
 
 	VM_BUG_ON(!IS_ENABLED(CONFIG_READ_ONLY_THP_FOR_FS) && !is_shmem);
 	VM_BUG_ON(start & (HPAGE_PMD_NR - 1));
 
 	/* Only allocate from the target node */
 	gfp = alloc_hugepage_khugepaged_gfpmask() | __GFP_THISNODE;
+	node = khugepaged_find_target_node(cc);
 
-	new_page = khugepaged_alloc_page(hpage, gfp, node);
+	new_page = cc->alloc_hpage(cc, gfp, node);
 	if (!new_page) {
 		result = SCAN_ALLOC_HUGE_PAGE_FAIL;
 		goto out;
@@ -1998,7 +2000,7 @@ static void collapse_file(struct mm_struct *mm,
 		 * Remove pte page tables, so we can re-fault the page as huge.
 		 */
 		retract_page_tables(mapping, start);
-		*hpage = NULL;
+		cc->hpage = NULL;
 
 		khugepaged_pages_collapsed++;
 	} else {
@@ -2045,14 +2047,14 @@ static void collapse_file(struct mm_struct *mm,
 	unlock_page(new_page);
 out:
 	VM_BUG_ON(!list_empty(&pagelist));
-	if (!IS_ERR_OR_NULL(*hpage))
-		mem_cgroup_uncharge(page_folio(*hpage));
+	if (!IS_ERR_OR_NULL(cc->hpage))
+		mem_cgroup_uncharge(page_folio(cc->hpage));
 	/* TODO: tracepoints */
 }
 
 static void khugepaged_scan_file(struct mm_struct *mm,
-		struct file *file, pgoff_t start, struct page **hpage,
-		struct collapse_control *cc)
+				 struct file *file, pgoff_t start,
+				 struct collapse_control *cc)
 {
 	struct page *page = NULL;
 	struct address_space *mapping = file->f_mapping;
@@ -2125,8 +2127,7 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 			result = SCAN_EXCEED_NONE_PTE;
 			count_vm_event(THP_SCAN_EXCEED_NONE_PTE);
 		} else {
-			node = khugepaged_find_target_node(cc);
-			collapse_file(mm, file, start, hpage, node);
+			collapse_file(mm, file, start, cc);
 		}
 	}
 
@@ -2134,8 +2135,8 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 }
 #else
 static void khugepaged_scan_file(struct mm_struct *mm,
-		struct file *file, pgoff_t start, struct page **hpage,
-		struct collapse_control *cc)
+				 struct file *file, pgoff_t start,
+				 struct collapse_control *cc)
 {
 	BUILD_BUG();
 }
@@ -2146,7 +2147,6 @@ static void khugepaged_collapse_pte_mapped_thps(struct mm_slot *mm_slot)
 #endif
 
 static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
-					    struct page **hpage,
 					    struct collapse_control *cc)
 	__releases(&khugepaged_mm_lock)
 	__acquires(&khugepaged_mm_lock)
@@ -2223,12 +2223,11 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
 
 				mmap_read_unlock(mm);
 				ret = 1;
-				khugepaged_scan_file(mm, file, pgoff, hpage, cc);
+				khugepaged_scan_file(mm, file, pgoff, cc);
 				fput(file);
 			} else {
 				ret = khugepaged_scan_pmd(mm, vma,
-						khugepaged_scan.address,
-						hpage, cc);
+						khugepaged_scan.address, cc);
 			}
 			/* move to next address */
 			khugepaged_scan.address += HPAGE_PMD_SIZE;
@@ -2286,15 +2285,15 @@ static int khugepaged_wait_event(void)
 
 static void khugepaged_do_scan(struct collapse_control *cc)
 {
-	struct page *hpage = NULL;
 	unsigned int progress = 0, pass_through_head = 0;
 	unsigned int pages = READ_ONCE(khugepaged_pages_to_scan);
 	bool wait = true;
 
+	cc->hpage = NULL;
 	lru_add_drain_all();
 
 	while (progress < pages) {
-		if (!khugepaged_prealloc_page(&hpage, &wait))
+		if (!khugepaged_prealloc_page(&cc->hpage, &wait))
 			break;
 
 		cond_resched();
@@ -2308,14 +2307,14 @@ static void khugepaged_do_scan(struct collapse_control *cc)
 		if (khugepaged_has_work() &&
 		    pass_through_head < 2)
 			progress += khugepaged_scan_mm_slot(pages - progress,
-							    &hpage, cc);
+							    cc);
 		else
 			progress = pages;
 		spin_unlock(&khugepaged_mm_lock);
 	}
 
-	if (!IS_ERR_OR_NULL(hpage))
-		put_page(hpage);
+	if (!IS_ERR_OR_NULL(cc->hpage))
+		put_page(cc->hpage);
 }
 
 static bool khugepaged_should_wakeup(void)
@@ -2349,6 +2348,7 @@ static int khugepaged(void *none)
 	struct mm_slot *mm_slot;
 	struct collapse_control cc = {
 		.last_target_node = NUMA_NO_NODE,
+		.alloc_hpage = &khugepaged_alloc_page,
 	};
 
 	set_freezable();
-- 
2.36.0.rc0.470.gd361397f0d-goog

[PATCH v2 04/12] mm/khugepaged: add struct collapse_result

[Zach O'Keefe]

Add struct collapse_result which aggregates data from a single
khugepaged_scan_pmd() or khugapaged_scan_file() request.  Change
khugepaged to take action based on this returned data instead of deep
within the collapsing functions themselves.

Signed-off-by: Zach O'Keefe <zokeefe@google.com>
---
 mm/khugepaged.c | 187 ++++++++++++++++++++++++++----------------------
 1 file changed, 101 insertions(+), 86 deletions(-)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 21c8436fa73c..e330a95a0479 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -98,6 +98,14 @@ struct collapse_control {
 				    int node);
 };
 
+/* Gather information from one khugepaged_scan_[pmd|file]() request */
+struct collapse_result {
+	enum scan_result result;
+
+	/* Was mmap_lock dropped during request? */
+	bool dropped_mmap_lock;
+};
+
 /**
  * struct mm_slot - hash lookup from mm to mm_slot
  * @hash: hash collision list
@@ -742,13 +750,13 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 		result = SCAN_SUCCEED;
 		trace_mm_collapse_huge_page_isolate(page, none_or_zero,
 						    referenced, writable, result);
-		return 1;
+		return SCAN_SUCCEED;
 	}
 out:
 	release_pte_pages(pte, _pte, compound_pagelist);
 	trace_mm_collapse_huge_page_isolate(page, none_or_zero,
 					    referenced, writable, result);
-	return 0;
+	return result;
 }
 
 static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
@@ -1086,7 +1094,7 @@ static bool __collapse_huge_page_swapin(struct mm_struct *mm,
 
 static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
 			       struct collapse_control *cc, int referenced,
-			       int unmapped)
+			       int unmapped, struct collapse_result *cr)
 {
 	LIST_HEAD(compound_pagelist);
 	pmd_t *pmd, _pmd;
@@ -1094,7 +1102,6 @@ static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
 	pgtable_t pgtable;
 	struct page *new_page;
 	spinlock_t *pmd_ptl, *pte_ptl;
-	int isolated = 0, result = 0;
 	struct vm_area_struct *vma;
 	struct mmu_notifier_range range;
 	gfp_t gfp;
@@ -1102,6 +1109,7 @@ static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
 	int node;
 
 	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
+	cr->result = SCAN_FAIL;
 
 	/* Only allocate from the target node */
 	gfp = alloc_hugepage_khugepaged_gfpmask() | __GFP_THISNODE;
@@ -1113,6 +1121,7 @@ static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
 	 * that. We will recheck the vma after taking it again in write mode.
 	 */
 	mmap_read_unlock(mm);
+	cr->dropped_mmap_lock = true;
 
 	node = khugepaged_find_target_node(cc);
 	/* sched to specified node before huage page memory copy */
@@ -1123,26 +1132,26 @@ static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
 	}
 	new_page = cc->alloc_hpage(cc, gfp, node);
 	if (!new_page) {
-		result = SCAN_ALLOC_HUGE_PAGE_FAIL;
+		cr->result = SCAN_ALLOC_HUGE_PAGE_FAIL;
 		goto out_nolock;
 	}
 
 	if (unlikely(mem_cgroup_charge(page_folio(new_page), mm, gfp))) {
-		result = SCAN_CGROUP_CHARGE_FAIL;
+		cr->result = SCAN_CGROUP_CHARGE_FAIL;
 		goto out_nolock;
 	}
 	count_memcg_page_event(new_page, THP_COLLAPSE_ALLOC);
 
 	mmap_read_lock(mm);
-	result = hugepage_vma_revalidate(mm, address, &vma);
-	if (result) {
+	cr->result = hugepage_vma_revalidate(mm, address, &vma);
+	if (cr->result) {
 		mmap_read_unlock(mm);
 		goto out_nolock;
 	}
 
 	pmd = mm_find_pmd(mm, address);
 	if (!pmd) {
-		result = SCAN_PMD_NULL;
+		cr->result = SCAN_PMD_NULL;
 		mmap_read_unlock(mm);
 		goto out_nolock;
 	}
@@ -1165,8 +1174,8 @@ static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
 	 * handled by the anon_vma lock + PG_lock.
 	 */
 	mmap_write_lock(mm);
-	result = hugepage_vma_revalidate(mm, address, &vma);
-	if (result)
+	cr->result = hugepage_vma_revalidate(mm, address, &vma);
+	if (cr->result)
 		goto out_up_write;
 	/* check if the pmd is still valid */
 	if (mm_find_pmd(mm, address) != pmd)
@@ -1193,11 +1202,11 @@ static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
 	mmu_notifier_invalidate_range_end(&range);
 
 	spin_lock(pte_ptl);
-	isolated = __collapse_huge_page_isolate(vma, address, pte,
-			&compound_pagelist);
+	cr->result =  __collapse_huge_page_isolate(vma, address, pte,
+						   &compound_pagelist);
 	spin_unlock(pte_ptl);
 
-	if (unlikely(!isolated)) {
+	if (unlikely(cr->result != SCAN_SUCCEED)) {
 		pte_unmap(pte);
 		spin_lock(pmd_ptl);
 		BUG_ON(!pmd_none(*pmd));
@@ -1209,7 +1218,7 @@ static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
 		pmd_populate(mm, pmd, pmd_pgtable(_pmd));
 		spin_unlock(pmd_ptl);
 		anon_vma_unlock_write(vma->anon_vma);
-		result = SCAN_FAIL;
+		cr->result = SCAN_FAIL;
 		goto out_up_write;
 	}
 
@@ -1245,25 +1254,25 @@ static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
 
 	cc->hpage = NULL;
 
-	khugepaged_pages_collapsed++;
-	result = SCAN_SUCCEED;
+	cr->result = SCAN_SUCCEED;
 out_up_write:
 	mmap_write_unlock(mm);
 out_nolock:
 	if (!IS_ERR_OR_NULL(cc->hpage))
 		mem_cgroup_uncharge(page_folio(cc->hpage));
-	trace_mm_collapse_huge_page(mm, isolated, result);
+	trace_mm_collapse_huge_page(mm, cr->result == SCAN_SUCCEED, cr->result);
 	return;
 }
 
-static int khugepaged_scan_pmd(struct mm_struct *mm,
-			       struct vm_area_struct *vma,
-			       unsigned long address,
-			       struct collapse_control *cc)
+static void khugepaged_scan_pmd(struct mm_struct *mm,
+				struct vm_area_struct *vma,
+				unsigned long address,
+				struct collapse_control *cc,
+				struct collapse_result *cr)
 {
 	pmd_t *pmd;
 	pte_t *pte, *_pte;
-	int ret = 0, result = 0, referenced = 0;
+	int referenced = 0;
 	int none_or_zero = 0, shared = 0;
 	struct page *page = NULL;
 	unsigned long _address;
@@ -1272,9 +1281,10 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 	bool writable = false;
 
 	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
+	cr->result = SCAN_FAIL;
 
-	result = find_pmd_or_thp_or_none(mm, address, &pmd);
-	if (result != SCAN_SUCCEED)
+	cr->result = find_pmd_or_thp_or_none(mm, address, &pmd);
+	if (cr->result != SCAN_SUCCEED)
 		goto out;
 
 	memset(cc->node_load, 0, sizeof(cc->node_load));
@@ -1290,12 +1300,12 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 				 * comment below for pte_uffd_wp().
 				 */
 				if (pte_swp_uffd_wp(pteval)) {
-					result = SCAN_PTE_UFFD_WP;
+					cr->result = SCAN_PTE_UFFD_WP;
 					goto out_unmap;
 				}
 				continue;
 			} else {
-				result = SCAN_EXCEED_SWAP_PTE;
+				cr->result = SCAN_EXCEED_SWAP_PTE;
 				count_vm_event(THP_SCAN_EXCEED_SWAP_PTE);
 				goto out_unmap;
 			}
@@ -1305,7 +1315,7 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 			    ++none_or_zero <= khugepaged_max_ptes_none) {
 				continue;
 			} else {
-				result = SCAN_EXCEED_NONE_PTE;
+				cr->result = SCAN_EXCEED_NONE_PTE;
 				count_vm_event(THP_SCAN_EXCEED_NONE_PTE);
 				goto out_unmap;
 			}
@@ -1320,7 +1330,7 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 			 * userfault messages that falls outside of
 			 * the registered range.  So, just be simple.
 			 */
-			result = SCAN_PTE_UFFD_WP;
+			cr->result = SCAN_PTE_UFFD_WP;
 			goto out_unmap;
 		}
 		if (pte_write(pteval))
@@ -1328,13 +1338,13 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 
 		page = vm_normal_page(vma, _address, pteval);
 		if (unlikely(!page)) {
-			result = SCAN_PAGE_NULL;
+			cr->result = SCAN_PAGE_NULL;
 			goto out_unmap;
 		}
 
 		if (page_mapcount(page) > 1 &&
 				++shared > khugepaged_max_ptes_shared) {
-			result = SCAN_EXCEED_SHARED_PTE;
+			cr->result = SCAN_EXCEED_SHARED_PTE;
 			count_vm_event(THP_SCAN_EXCEED_SHARED_PTE);
 			goto out_unmap;
 		}
@@ -1349,20 +1359,20 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 		 */
 		node = page_to_nid(page);
 		if (khugepaged_scan_abort(node, cc)) {
-			result = SCAN_SCAN_ABORT;
+			cr->result = SCAN_SCAN_ABORT;
 			goto out_unmap;
 		}
 		cc->node_load[node]++;
 		if (!PageLRU(page)) {
-			result = SCAN_PAGE_LRU;
+			cr->result = SCAN_PAGE_LRU;
 			goto out_unmap;
 		}
 		if (PageLocked(page)) {
-			result = SCAN_PAGE_LOCK;
+			cr->result = SCAN_PAGE_LOCK;
 			goto out_unmap;
 		}
 		if (!PageAnon(page)) {
-			result = SCAN_PAGE_ANON;
+			cr->result = SCAN_PAGE_ANON;
 			goto out_unmap;
 		}
 
@@ -1384,7 +1394,7 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 		 * will be done again later the risk seems low.
 		 */
 		if (!is_refcount_suitable(page)) {
-			result = SCAN_PAGE_COUNT;
+			cr->result = SCAN_PAGE_COUNT;
 			goto out_unmap;
 		}
 		if (pte_young(pteval) ||
@@ -1393,23 +1403,20 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
 			referenced++;
 	}
 	if (!writable) {
-		result = SCAN_PAGE_RO;
+		cr->result = SCAN_PAGE_RO;
 	} else if (!referenced || (unmapped && referenced < HPAGE_PMD_NR/2)) {
-		result = SCAN_LACK_REFERENCED_PAGE;
+		cr->result = SCAN_LACK_REFERENCED_PAGE;
 	} else {
-		result = SCAN_SUCCEED;
-		ret = 1;
+		cr->result = SCAN_SUCCEED;
 	}
 out_unmap:
 	pte_unmap_unlock(pte, ptl);
-	if (ret) {
+	if (cr->result == SCAN_SUCCEED)
 		/* collapse_huge_page will return with the mmap_lock released */
-		collapse_huge_page(mm, address, cc, referenced, unmapped);
-	}
+		collapse_huge_page(mm, address, cc, referenced, unmapped, cr);
 out:
 	trace_mm_khugepaged_scan_pmd(mm, page, writable, referenced,
-				     none_or_zero, result, unmapped);
-	return ret;
+				     none_or_zero, cr->result, unmapped);
 }
 
 static void collect_mm_slot(struct mm_slot *mm_slot)
@@ -1670,6 +1677,7 @@ static void retract_page_tables(struct address_space *mapping, pgoff_t pgoff)
  * @file: file that collapse on
  * @start: collapse start address
  * @cc: collapse context and scratchpad
+ * @cr: aggregate result information of collapse
  *
  * Basic scheme is simple, details are more complex:
  *  - allocate and lock a new huge page;
@@ -1688,7 +1696,9 @@ static void retract_page_tables(struct address_space *mapping, pgoff_t pgoff)
  */
 static void collapse_file(struct mm_struct *mm,
 			  struct file *file, pgoff_t start,
-			  struct collapse_control *cc)
+			  struct collapse_control *cc,
+			  struct collapse_result *cr)
+
 {
 	struct address_space *mapping = file->f_mapping;
 	gfp_t gfp;
@@ -1696,25 +1706,27 @@ static void collapse_file(struct mm_struct *mm,
 	pgoff_t index, end = start + HPAGE_PMD_NR;
 	LIST_HEAD(pagelist);
 	XA_STATE_ORDER(xas, &mapping->i_pages, start, HPAGE_PMD_ORDER);
-	int nr_none = 0, result = SCAN_SUCCEED;
+	int nr_none = 0;
 	bool is_shmem = shmem_file(file);
 	int nr, node;
 
 	VM_BUG_ON(!IS_ENABLED(CONFIG_READ_ONLY_THP_FOR_FS) && !is_shmem);
 	VM_BUG_ON(start & (HPAGE_PMD_NR - 1));
 
+	cr->result = SCAN_SUCCEED;
+
 	/* Only allocate from the target node */
 	gfp = alloc_hugepage_khugepaged_gfpmask() | __GFP_THISNODE;
 	node = khugepaged_find_target_node(cc);
 
 	new_page = cc->alloc_hpage(cc, gfp, node);
 	if (!new_page) {
-		result = SCAN_ALLOC_HUGE_PAGE_FAIL;
+		cr->result = SCAN_ALLOC_HUGE_PAGE_FAIL;
 		goto out;
 	}
 
 	if (unlikely(mem_cgroup_charge(page_folio(new_page), mm, gfp))) {
-		result = SCAN_CGROUP_CHARGE_FAIL;
+		cr->result = SCAN_CGROUP_CHARGE_FAIL;
 		goto out;
 	}
 	count_memcg_page_event(new_page, THP_COLLAPSE_ALLOC);
@@ -1730,7 +1742,7 @@ static void collapse_file(struct mm_struct *mm,
 			break;
 		xas_unlock_irq(&xas);
 		if (!xas_nomem(&xas, GFP_KERNEL)) {
-			result = SCAN_FAIL;
+			cr->result = SCAN_FAIL;
 			goto out;
 		}
 	} while (1);
@@ -1761,13 +1773,13 @@ static void collapse_file(struct mm_struct *mm,
 				 */
 				if (index == start) {
 					if (!xas_next_entry(&xas, end - 1)) {
-						result = SCAN_TRUNCATED;
+						cr->result = SCAN_TRUNCATED;
 						goto xa_locked;
 					}
 					xas_set(&xas, index);
 				}
 				if (!shmem_charge(mapping->host, 1)) {
-					result = SCAN_FAIL;
+					cr->result = SCAN_FAIL;
 					goto xa_locked;
 				}
 				xas_store(&xas, new_page);
@@ -1780,14 +1792,14 @@ static void collapse_file(struct mm_struct *mm,
 				/* swap in or instantiate fallocated page */
 				if (shmem_getpage(mapping->host, index, &page,
 						  SGP_NOALLOC)) {
-					result = SCAN_FAIL;
+					cr->result = SCAN_FAIL;
 					goto xa_unlocked;
 				}
 			} else if (trylock_page(page)) {
 				get_page(page);
 				xas_unlock_irq(&xas);
 			} else {
-				result = SCAN_PAGE_LOCK;
+				cr->result = SCAN_PAGE_LOCK;
 				goto xa_locked;
 			}
 		} else {	/* !is_shmem */
@@ -1800,7 +1812,7 @@ static void collapse_file(struct mm_struct *mm,
 				lru_add_drain();
 				page = find_lock_page(mapping, index);
 				if (unlikely(page == NULL)) {
-					result = SCAN_FAIL;
+					cr->result = SCAN_FAIL;
 					goto xa_unlocked;
 				}
 			} else if (PageDirty(page)) {
@@ -1819,17 +1831,17 @@ static void collapse_file(struct mm_struct *mm,
 				 */
 				xas_unlock_irq(&xas);
 				filemap_flush(mapping);
-				result = SCAN_FAIL;
+				cr->result = SCAN_FAIL;
 				goto xa_unlocked;
 			} else if (PageWriteback(page)) {
 				xas_unlock_irq(&xas);
-				result = SCAN_FAIL;
+				cr->result = SCAN_FAIL;
 				goto xa_unlocked;
 			} else if (trylock_page(page)) {
 				get_page(page);
 				xas_unlock_irq(&xas);
 			} else {
-				result = SCAN_PAGE_LOCK;
+				cr->result = SCAN_PAGE_LOCK;
 				goto xa_locked;
 			}
 		}
@@ -1842,7 +1854,7 @@ static void collapse_file(struct mm_struct *mm,
 
 		/* make sure the page is up to date */
 		if (unlikely(!PageUptodate(page))) {
-			result = SCAN_FAIL;
+			cr->result = SCAN_FAIL;
 			goto out_unlock;
 		}
 
@@ -1851,12 +1863,12 @@ static void collapse_file(struct mm_struct *mm,
 		 * we locked the first page, then a THP might be there already.
 		 */
 		if (PageTransCompound(page)) {
-			result = SCAN_PAGE_COMPOUND;
+			cr->result = SCAN_PAGE_COMPOUND;
 			goto out_unlock;
 		}
 
 		if (page_mapping(page) != mapping) {
-			result = SCAN_TRUNCATED;
+			cr->result = SCAN_TRUNCATED;
 			goto out_unlock;
 		}
 
@@ -1867,18 +1879,18 @@ static void collapse_file(struct mm_struct *mm,
 			 * page is dirty because it hasn't been flushed
 			 * since first write.
 			 */
-			result = SCAN_FAIL;
+			cr->result = SCAN_FAIL;
 			goto out_unlock;
 		}
 
 		if (isolate_lru_page(page)) {
-			result = SCAN_DEL_PAGE_LRU;
+			cr->result = SCAN_DEL_PAGE_LRU;
 			goto out_unlock;
 		}
 
 		if (page_has_private(page) &&
 		    !try_to_release_page(page, GFP_KERNEL)) {
-			result = SCAN_PAGE_HAS_PRIVATE;
+			cr->result = SCAN_PAGE_HAS_PRIVATE;
 			putback_lru_page(page);
 			goto out_unlock;
 		}
@@ -1899,7 +1911,7 @@ static void collapse_file(struct mm_struct *mm,
 		 *  - one from isolate_lru_page;
 		 */
 		if (!page_ref_freeze(page, 3)) {
-			result = SCAN_PAGE_COUNT;
+			cr->result = SCAN_PAGE_COUNT;
 			xas_unlock_irq(&xas);
 			putback_lru_page(page);
 			goto out_unlock;
@@ -1934,7 +1946,7 @@ static void collapse_file(struct mm_struct *mm,
 		 */
 		smp_mb();
 		if (inode_is_open_for_write(mapping->host)) {
-			result = SCAN_FAIL;
+			cr->result = SCAN_FAIL;
 			__mod_lruvec_page_state(new_page, NR_FILE_THPS, -nr);
 			filemap_nr_thps_dec(mapping);
 			goto xa_locked;
@@ -1961,7 +1973,7 @@ static void collapse_file(struct mm_struct *mm,
 	 */
 	try_to_unmap_flush();
 
-	if (result == SCAN_SUCCEED) {
+	if (cr->result == SCAN_SUCCEED) {
 		struct page *page, *tmp;
 
 		/*
@@ -2001,8 +2013,6 @@ static void collapse_file(struct mm_struct *mm,
 		 */
 		retract_page_tables(mapping, start);
 		cc->hpage = NULL;
-
-		khugepaged_pages_collapsed++;
 	} else {
 		struct page *page;
 
@@ -2054,15 +2064,16 @@ static void collapse_file(struct mm_struct *mm,
 
 static void khugepaged_scan_file(struct mm_struct *mm,
 				 struct file *file, pgoff_t start,
-				 struct collapse_control *cc)
+				 struct collapse_control *cc,
+				 struct collapse_result *cr)
 {
 	struct page *page = NULL;
 	struct address_space *mapping = file->f_mapping;
 	XA_STATE(xas, &mapping->i_pages, start);
 	int present, swap;
 	int node = NUMA_NO_NODE;
-	int result = SCAN_SUCCEED;
 
+	cr->result = SCAN_SUCCEED;
 	present = 0;
 	swap = 0;
 	memset(cc->node_load, 0, sizeof(cc->node_load));
@@ -2073,7 +2084,7 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 
 		if (xa_is_value(page)) {
 			if (++swap > khugepaged_max_ptes_swap) {
-				result = SCAN_EXCEED_SWAP_PTE;
+				cr->result = SCAN_EXCEED_SWAP_PTE;
 				count_vm_event(THP_SCAN_EXCEED_SWAP_PTE);
 				break;
 			}
@@ -2085,25 +2096,25 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 		 * into a PMD sized page
 		 */
 		if (PageTransCompound(page)) {
-			result = SCAN_PAGE_COMPOUND;
+			cr->result = SCAN_PAGE_COMPOUND;
 			break;
 		}
 
 		node = page_to_nid(page);
 		if (khugepaged_scan_abort(node, cc)) {
-			result = SCAN_SCAN_ABORT;
+			cr->result = SCAN_SCAN_ABORT;
 			break;
 		}
 		cc->node_load[node]++;
 
 		if (!PageLRU(page)) {
-			result = SCAN_PAGE_LRU;
+			cr->result = SCAN_PAGE_LRU;
 			break;
 		}
 
 		if (page_count(page) !=
 		    1 + page_mapcount(page) + page_has_private(page)) {
-			result = SCAN_PAGE_COUNT;
+			cr->result = SCAN_PAGE_COUNT;
 			break;
 		}
 
@@ -2122,12 +2133,12 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 	}
 	rcu_read_unlock();
 
-	if (result == SCAN_SUCCEED) {
+	if (cr->result == SCAN_SUCCEED) {
 		if (present < HPAGE_PMD_NR - khugepaged_max_ptes_none) {
-			result = SCAN_EXCEED_NONE_PTE;
+			cr->result = SCAN_EXCEED_NONE_PTE;
 			count_vm_event(THP_SCAN_EXCEED_NONE_PTE);
 		} else {
-			collapse_file(mm, file, start, cc);
+			collapse_file(mm, file, start, cc, cr);
 		}
 	}
 
@@ -2136,7 +2147,8 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 #else
 static void khugepaged_scan_file(struct mm_struct *mm,
 				 struct file *file, pgoff_t start,
-				 struct collapse_control *cc)
+				 struct collapse_control *cc,
+				 struct collapse_result *cr)
 {
 	BUILD_BUG();
 }
@@ -2208,7 +2220,7 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
 			goto skip;
 
 		while (khugepaged_scan.address < hend) {
-			int ret;
+			struct collapse_result cr = {0};
 			cond_resched();
 			if (unlikely(khugepaged_test_exit(mm)))
 				goto breakouterloop;
@@ -2222,17 +2234,20 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
 						khugepaged_scan.address);
 
 				mmap_read_unlock(mm);
-				ret = 1;
-				khugepaged_scan_file(mm, file, pgoff, cc);
+				cr.dropped_mmap_lock = true;
+				khugepaged_scan_file(mm, file, pgoff, cc, &cr);
 				fput(file);
 			} else {
-				ret = khugepaged_scan_pmd(mm, vma,
-						khugepaged_scan.address, cc);
+				khugepaged_scan_pmd(mm, vma,
+						    khugepaged_scan.address,
+						    cc, &cr);
 			}
+			if (cr.result == SCAN_SUCCEED)
+				++khugepaged_pages_collapsed;
 			/* move to next address */
 			khugepaged_scan.address += HPAGE_PMD_SIZE;
 			progress += HPAGE_PMD_NR;
-			if (ret)
+			if (cr.dropped_mmap_lock)
 				/* we released mmap_lock so break loop */
 				goto breakouterloop_mmap_lock;
 			if (progress >= pages)
-- 
2.36.0.rc0.470.gd361397f0d-goog

[PATCH v2 05/12] mm/madvise: introduce MADV_COLLAPSE sync hugepage collapse

[Zach O'Keefe]

This idea was introduced by David Rientjes[1], and the semantics and
implementation were introduced and discussed in a previous PATCH RFC[2].

Introduce a new madvise mode, MADV_COLLAPSE, that allows users to request a
synchronous collapse of memory at their own expense.

The benefits of this approach are:

* CPU is charged to the process that wants to spend the cycles for the
  THP
* avoid unpredictable timing of khugepaged collapse

Immediate users of this new functionality include:

* immediately back executable text by hugepages.  Current support
  provided by CONFIG_READ_ONLY_THP_FOR_FS may take too long on a large
  system.
* malloc implementations that manage memory in hugepage-sized chunks,
  but sometimes subrelease memory back to the system in native-sized
  chunks via MADV_DONTNEED; zapping the pmd.  Later, when the memory
  is hot, the implementation could madvise(MADV_COLLAPSE) to re-back the
  memory by THP to regain TLB performance.

Allocation semantics are the same as khugepaged, and depend on (1) the
active sysfs settings /sys/kernel/mm/transparent_hugepage/enabled and
/sys/kernel/mm/transparent_hugepage/khugepaged/defrag, and (2) the VMA
flags of the memory range being collapsed.

Only privately-mapped anon memory is supported for now.

[1] https://lore.kernel.org/linux-mm/d098c392-273a-36a4-1a29-59731cdf5d3d@google.com/
[2] https://lore.kernel.org/linux-mm/20220308213417.1407042-1-zokeefe@google.com/

Suggested-by: David Rientjes <rientjes@google.com>
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reported-by: kernel test robot <lkp@intel.com>
---
 arch/alpha/include/uapi/asm/mman.h     |   2 +
 arch/mips/include/uapi/asm/mman.h      |   2 +
 arch/parisc/include/uapi/asm/mman.h    |   2 +
 arch/xtensa/include/uapi/asm/mman.h    |   2 +
 include/linux/huge_mm.h                |  12 ++
 include/uapi/asm-generic/mman-common.h |   2 +
 mm/khugepaged.c                        | 149 +++++++++++++++++++++++--
 mm/madvise.c                           |   5 +
 8 files changed, 164 insertions(+), 12 deletions(-)

diff --git a/arch/alpha/include/uapi/asm/mman.h b/arch/alpha/include/uapi/asm/mman.h
index 4aa996423b0d..763929e814e9 100644
--- a/arch/alpha/include/uapi/asm/mman.h
+++ b/arch/alpha/include/uapi/asm/mman.h
@@ -76,6 +76,8 @@
 
 #define MADV_DONTNEED_LOCKED	24	/* like DONTNEED, but drop locked pages too */
 
+#define MADV_COLLAPSE	25		/* Synchronous hugepage collapse */
+
 /* compatibility flags */
 #define MAP_FILE	0
 
diff --git a/arch/mips/include/uapi/asm/mman.h b/arch/mips/include/uapi/asm/mman.h
index 1be428663c10..c6e1fc77c996 100644
--- a/arch/mips/include/uapi/asm/mman.h
+++ b/arch/mips/include/uapi/asm/mman.h
@@ -103,6 +103,8 @@
 
 #define MADV_DONTNEED_LOCKED	24	/* like DONTNEED, but drop locked pages too */
 
+#define MADV_COLLAPSE	25		/* Synchronous hugepage collapse */
+
 /* compatibility flags */
 #define MAP_FILE	0
 
diff --git a/arch/parisc/include/uapi/asm/mman.h b/arch/parisc/include/uapi/asm/mman.h
index a7ea3204a5fa..22133a6a506e 100644
--- a/arch/parisc/include/uapi/asm/mman.h
+++ b/arch/parisc/include/uapi/asm/mman.h
@@ -70,6 +70,8 @@
 #define MADV_WIPEONFORK 71		/* Zero memory on fork, child only */
 #define MADV_KEEPONFORK 72		/* Undo MADV_WIPEONFORK */
 
+#define MADV_COLLAPSE	73		/* Synchronous hugepage collapse */
+
 #define MADV_HWPOISON     100		/* poison a page for testing */
 #define MADV_SOFT_OFFLINE 101		/* soft offline page for testing */
 
diff --git a/arch/xtensa/include/uapi/asm/mman.h b/arch/xtensa/include/uapi/asm/mman.h
index 7966a58af472..1ff0c858544f 100644
--- a/arch/xtensa/include/uapi/asm/mman.h
+++ b/arch/xtensa/include/uapi/asm/mman.h
@@ -111,6 +111,8 @@
 
 #define MADV_DONTNEED_LOCKED	24	/* like DONTNEED, but drop locked pages too */
 
+#define MADV_COLLAPSE	25		/* Synchronous hugepage collapse */
+
 /* compatibility flags */
 #define MAP_FILE	0
 
diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
index 816a9937f30e..ddad7c7af44e 100644
--- a/include/linux/huge_mm.h
+++ b/include/linux/huge_mm.h
@@ -236,6 +236,9 @@ void __split_huge_pud(struct vm_area_struct *vma, pud_t *pud,
 
 int hugepage_madvise(struct vm_area_struct *vma, unsigned long *vm_flags,
 		     int advice);
+int madvise_collapse(struct vm_area_struct *vma,
+		     struct vm_area_struct **prev,
+		     unsigned long start, unsigned long end);
 void vma_adjust_trans_huge(struct vm_area_struct *vma, unsigned long start,
 			   unsigned long end, long adjust_next);
 spinlock_t *__pmd_trans_huge_lock(pmd_t *pmd, struct vm_area_struct *vma);
@@ -392,6 +395,15 @@ static inline int hugepage_madvise(struct vm_area_struct *vma,
 	BUG();
 	return 0;
 }
+
+static inline int madvise_collapse(struct vm_area_struct *vma,
+				   struct vm_area_struct **prev,
+				   unsigned long start, unsigned long end)
+{
+	BUG();
+	return 0;
+}
+
 static inline void vma_adjust_trans_huge(struct vm_area_struct *vma,
 					 unsigned long start,
 					 unsigned long end,
diff --git a/include/uapi/asm-generic/mman-common.h b/include/uapi/asm-generic/mman-common.h
index 6c1aa92a92e4..6ce1f1ceb432 100644
--- a/include/uapi/asm-generic/mman-common.h
+++ b/include/uapi/asm-generic/mman-common.h
@@ -77,6 +77,8 @@
 
 #define MADV_DONTNEED_LOCKED	24	/* like DONTNEED, but drop locked pages too */
 
+#define MADV_COLLAPSE	25		/* Synchronous hugepage collapse */
+
 /* compatibility flags */
 #define MAP_FILE	0
 
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index e330a95a0479..c745829c3965 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -846,6 +846,23 @@ static inline gfp_t alloc_hugepage_khugepaged_gfpmask(void)
 	return khugepaged_defrag() ? GFP_TRANSHUGE : GFP_TRANSHUGE_LIGHT;
 }
 
+static struct page *alloc_hpage(struct collapse_control *cc, gfp_t gfp,
+				int node)
+{
+	VM_BUG_ON_PAGE(cc->hpage, cc->hpage);
+
+	cc->hpage = __alloc_pages_node(node, gfp, HPAGE_PMD_ORDER);
+	if (unlikely(!cc->hpage)) {
+		count_vm_event(THP_COLLAPSE_ALLOC_FAILED);
+		cc->hpage = ERR_PTR(-ENOMEM);
+		return NULL;
+	}
+
+	prep_transhuge_page(cc->hpage);
+	count_vm_event(THP_COLLAPSE_ALLOC);
+	return cc->hpage;
+}
+
 #ifdef CONFIG_NUMA
 static int khugepaged_find_target_node(struct collapse_control *cc)
 {
@@ -892,18 +909,7 @@ static bool khugepaged_prealloc_page(struct page **hpage, bool *wait)
 static struct page *khugepaged_alloc_page(struct collapse_control *cc,
 					  gfp_t gfp, int node)
 {
-	VM_BUG_ON_PAGE(cc->hpage, cc->hpage);
-
-	cc->hpage = __alloc_pages_node(node, gfp, HPAGE_PMD_ORDER);
-	if (unlikely(!cc->hpage)) {
-		count_vm_event(THP_COLLAPSE_ALLOC_FAILED);
-		cc->hpage = ERR_PTR(-ENOMEM);
-		return NULL;
-	}
-
-	prep_transhuge_page(cc->hpage);
-	count_vm_event(THP_COLLAPSE_ALLOC);
-	return cc->hpage;
+	return alloc_hpage(cc, gfp, node);
 }
 #else
 static int khugepaged_find_target_node(struct collapse_control *cc)
@@ -2469,3 +2475,122 @@ void khugepaged_min_free_kbytes_update(void)
 		set_recommended_min_free_kbytes();
 	mutex_unlock(&khugepaged_mutex);
 }
+
+static void madvise_collapse_cleanup_page(struct page **hpage)
+{
+	if (!IS_ERR(*hpage) && *hpage)
+		put_page(*hpage);
+	*hpage = NULL;
+}
+
+static int madvise_collapse_errno(enum scan_result r)
+{
+	switch (r) {
+	case SCAN_PMD_NULL:
+	case SCAN_ADDRESS_RANGE:
+	case SCAN_VMA_NULL:
+	case SCAN_PTE_NON_PRESENT:
+	case SCAN_PAGE_NULL:
+		/*
+		 * Addresses in the specified range are not currently mapped,
+		 * or are outside the AS of the process.
+		 */
+		return -ENOMEM;
+	case SCAN_ALLOC_HUGE_PAGE_FAIL:
+	case SCAN_CGROUP_CHARGE_FAIL:
+		/* A kernel resource was temporarily unavailable. */
+		return -EAGAIN;
+	default:
+		return -EINVAL;
+	}
+}
+
+int madvise_collapse(struct vm_area_struct *vma, struct vm_area_struct **prev,
+		     unsigned long start, unsigned long end)
+{
+	struct collapse_control cc = {
+		.last_target_node = NUMA_NO_NODE,
+		.hpage = NULL,
+		.alloc_hpage = &alloc_hpage,
+	};
+	struct mm_struct *mm = vma->vm_mm;
+	struct collapse_result cr;
+	unsigned long hstart, hend, addr;
+	int thps = 0, nr_hpages = 0;
+
+	BUG_ON(vma->vm_start > start);
+	BUG_ON(vma->vm_end < end);
+
+	*prev = vma;
+
+	if (IS_ENABLED(CONFIG_SHMEM) && vma->vm_file)
+		return -EINVAL;
+
+	hstart = (start + ~HPAGE_PMD_MASK) & HPAGE_PMD_MASK;
+	hend = end & HPAGE_PMD_MASK;
+	nr_hpages = (hend - hstart) >> HPAGE_PMD_SHIFT;
+
+	if (hstart >= hend || !transparent_hugepage_active(vma))
+		return -EINVAL;
+
+	mmgrab(mm);
+	lru_add_drain();
+
+	for (addr = hstart; ; ) {
+		mmap_assert_locked(mm);
+		cond_resched();
+		memset(&cr, 0, sizeof(cr));
+
+		if (unlikely(khugepaged_test_exit(mm)))
+			break;
+
+		memset(cc.node_load, 0, sizeof(cc.node_load));
+		khugepaged_scan_pmd(mm, vma, addr, &cc, &cr);
+		if (cr.dropped_mmap_lock)
+			*prev = NULL;  /* tell madvise we dropped mmap_lock */
+
+		switch (cr.result) {
+		/* Whitelisted set of results where continuing OK */
+		case SCAN_SUCCEED:
+		case SCAN_PMD_MAPPED:
+			++thps;
+		case SCAN_PMD_NULL:
+		case SCAN_PTE_NON_PRESENT:
+		case SCAN_PTE_UFFD_WP:
+		case SCAN_PAGE_RO:
+		case SCAN_LACK_REFERENCED_PAGE:
+		case SCAN_PAGE_NULL:
+		case SCAN_PAGE_COUNT:
+		case SCAN_PAGE_LOCK:
+		case SCAN_PAGE_COMPOUND:
+			break;
+		case SCAN_PAGE_LRU:
+			lru_add_drain_all();
+			goto retry;
+		default:
+			/* Other error, exit */
+			goto break_loop;
+		}
+		addr += HPAGE_PMD_SIZE;
+		if (addr >= hend)
+			break;
+retry:
+		if (cr.dropped_mmap_lock) {
+			mmap_read_lock(mm);
+			if (hugepage_vma_revalidate(mm, addr, &vma))
+				goto out;
+		}
+		madvise_collapse_cleanup_page(&cc.hpage);
+	}
+
+break_loop:
+	/* madvise_walk_vmas() expects us to hold mmap_lock on return */
+	if (cr.dropped_mmap_lock)
+		mmap_read_lock(mm);
+out:
+	mmap_assert_locked(mm);
+	madvise_collapse_cleanup_page(&cc.hpage);
+	mmdrop(mm);
+
+	return thps == nr_hpages ? 0 : madvise_collapse_errno(cr.result);
+}
diff --git a/mm/madvise.c b/mm/madvise.c
index ec03a76244b7..7ad53e5311cf 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -59,6 +59,7 @@ static int madvise_need_mmap_write(int behavior)
 	case MADV_FREE:
 	case MADV_POPULATE_READ:
 	case MADV_POPULATE_WRITE:
+	case MADV_COLLAPSE:
 		return 0;
 	default:
 		/* be safe, default to 1. list exceptions explicitly */
@@ -1051,6 +1052,8 @@ static int madvise_vma_behavior(struct vm_area_struct *vma,
 		if (error)
 			goto out;
 		break;
+	case MADV_COLLAPSE:
+		return madvise_collapse(vma, prev, start, end);
 	}
 
 	anon_name = anon_vma_name(vma);
@@ -1144,6 +1147,7 @@ madvise_behavior_valid(int behavior)
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 	case MADV_HUGEPAGE:
 	case MADV_NOHUGEPAGE:
+	case MADV_COLLAPSE:
 #endif
 	case MADV_DONTDUMP:
 	case MADV_DODUMP:
@@ -1333,6 +1337,7 @@ int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
  *  MADV_NOHUGEPAGE - mark the given range as not worth being backed by
  *		transparent huge pages so the existing pages will not be
  *		coalesced into THP and new pages will not be allocated as THP.
+ *  MADV_COLLAPSE - synchronously coalesce pages into new THP.
  *  MADV_DONTDUMP - the application wants to prevent pages in the given range
  *		from being included in its core dump.
  *  MADV_DODUMP - cancel MADV_DONTDUMP: no longer exclude from core dump.
-- 
2.36.0.rc0.470.gd361397f0d-goog

[PATCH v2 06/12] mm/khugepaged: remove khugepaged prefix from shared collapse functions

[Zach O'Keefe]

The following functions/tracepoints are shared between khugepaged and
madvise collapse contexts.  Remove the khugepaged prefixes.

tracepoint:mm_khugepaged_scan_pmd -> tracepoint:mm_scan_pmd
khugepaged_test_exit() -> test_exit()
khugepaged_scan_abort() -> scan_abort()
khugepaged_scan_pmd() -> scan_pmd()
khugepaged_find_target_node() -> find_target_node()

Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reported-by: kernel test robot <lkp@intel.com>
---
 include/trace/events/huge_memory.h |  2 +-
 mm/khugepaged.c                    | 70 ++++++++++++++----------------
 2 files changed, 34 insertions(+), 38 deletions(-)

diff --git a/include/trace/events/huge_memory.h b/include/trace/events/huge_memory.h
index 9faa678e0a5b..09be0e2f76b1 100644
--- a/include/trace/events/huge_memory.h
+++ b/include/trace/events/huge_memory.h
@@ -48,7 +48,7 @@ SCAN_STATUS
 #define EM(a, b)	{a, b},
 #define EMe(a, b)	{a, b}
 
-TRACE_EVENT(mm_khugepaged_scan_pmd,
+TRACE_EVENT(mm_scan_pmd,
 
 	TP_PROTO(struct mm_struct *mm, struct page *page, bool writable,
 		 int referenced, int none_or_zero, int status, int unmapped),
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index c745829c3965..716ba465b356 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -90,7 +90,7 @@ struct collapse_control {
 	/* Num pages scanned per node */
 	int node_load[MAX_NUMNODES];
 
-	/* Last target selected in khugepaged_find_target_node() for this scan */
+	/* Last target selected in find_target_node() for this scan */
 	int last_target_node;
 
 	struct page *hpage;
@@ -453,7 +453,7 @@ static void insert_to_mm_slots_hash(struct mm_struct *mm,
 	hash_add(mm_slots_hash, &mm_slot->hash, (long)mm);
 }
 
-static inline int khugepaged_test_exit(struct mm_struct *mm)
+static inline int test_exit(struct mm_struct *mm)
 {
 	return atomic_read(&mm->mm_users) == 0;
 }
@@ -505,7 +505,7 @@ void __khugepaged_enter(struct mm_struct *mm)
 		return;
 
 	/* __khugepaged_exit() must not run from under us */
-	VM_BUG_ON_MM(khugepaged_test_exit(mm), mm);
+	VM_BUG_ON_MM(test_exit(mm), mm);
 	if (unlikely(test_and_set_bit(MMF_VM_HUGEPAGE, &mm->flags))) {
 		free_mm_slot(mm_slot);
 		return;
@@ -557,12 +557,11 @@ void __khugepaged_exit(struct mm_struct *mm)
 		mmdrop(mm);
 	} else if (mm_slot) {
 		/*
-		 * This is required to serialize against
-		 * khugepaged_test_exit() (which is guaranteed to run
-		 * under mmap sem read mode). Stop here (after we
-		 * return all pagetables will be destroyed) until
-		 * khugepaged has finished working on the pagetables
-		 * under the mmap_lock.
+		 * This is required to serialize against test_exit() (which is
+		 * guaranteed to run under mmap sem read mode). Stop here
+		 * (after we return all pagetables will be destroyed) until
+		 * khugepaged has finished working on the pagetables under
+		 * the mmap_lock.
 		 */
 		mmap_write_lock(mm);
 		mmap_write_unlock(mm);
@@ -816,7 +815,7 @@ static void khugepaged_alloc_sleep(void)
 	remove_wait_queue(&khugepaged_wait, &wait);
 }
 
-static bool khugepaged_scan_abort(int nid, struct collapse_control *cc)
+static bool scan_abort(int nid, struct collapse_control *cc)
 {
 	int i;
 
@@ -864,7 +863,7 @@ static struct page *alloc_hpage(struct collapse_control *cc, gfp_t gfp,
 }
 
 #ifdef CONFIG_NUMA
-static int khugepaged_find_target_node(struct collapse_control *cc)
+static int find_target_node(struct collapse_control *cc)
 {
 	int nid, target_node = 0, max_value = 0;
 
@@ -912,7 +911,7 @@ static struct page *khugepaged_alloc_page(struct collapse_control *cc,
 	return alloc_hpage(cc, gfp, node);
 }
 #else
-static int khugepaged_find_target_node(struct collapse_control *cc)
+static int find_target_node(struct collapse_control *cc)
 {
 	return 0;
 }
@@ -993,7 +992,7 @@ static int hugepage_vma_revalidate(struct mm_struct *mm, unsigned long address,
 	struct vm_area_struct *vma;
 	unsigned long hstart, hend;
 
-	if (unlikely(khugepaged_test_exit(mm)))
+	if (unlikely(test_exit(mm)))
 		return SCAN_ANY_PROCESS;
 
 	*vmap = vma = find_vma(mm, address);
@@ -1037,7 +1036,7 @@ static int find_pmd_or_thp_or_none(struct mm_struct *mm,
 
 /*
  * Bring missing pages in from swap, to complete THP collapse.
- * Only done if khugepaged_scan_pmd believes it is worthwhile.
+ * Only done if scan_pmd believes it is worthwhile.
  *
  * Called and returns without pte mapped or spinlocks held,
  * but with mmap_lock held to protect against vma changes.
@@ -1129,7 +1128,7 @@ static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
 	mmap_read_unlock(mm);
 	cr->dropped_mmap_lock = true;
 
-	node = khugepaged_find_target_node(cc);
+	node = find_target_node(cc);
 	/* sched to specified node before huage page memory copy */
 	if (task_node(current) != node) {
 		cpumask = cpumask_of_node(node);
@@ -1270,11 +1269,9 @@ static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
 	return;
 }
 
-static void khugepaged_scan_pmd(struct mm_struct *mm,
-				struct vm_area_struct *vma,
-				unsigned long address,
-				struct collapse_control *cc,
-				struct collapse_result *cr)
+static void scan_pmd(struct mm_struct *mm, struct vm_area_struct *vma,
+		     unsigned long address, struct collapse_control *cc,
+		     struct collapse_result *cr)
 {
 	pmd_t *pmd;
 	pte_t *pte, *_pte;
@@ -1364,7 +1361,7 @@ static void khugepaged_scan_pmd(struct mm_struct *mm,
 		 * hit record.
 		 */
 		node = page_to_nid(page);
-		if (khugepaged_scan_abort(node, cc)) {
+		if (scan_abort(node, cc)) {
 			cr->result = SCAN_SCAN_ABORT;
 			goto out_unmap;
 		}
@@ -1421,8 +1418,8 @@ static void khugepaged_scan_pmd(struct mm_struct *mm,
 		/* collapse_huge_page will return with the mmap_lock released */
 		collapse_huge_page(mm, address, cc, referenced, unmapped, cr);
 out:
-	trace_mm_khugepaged_scan_pmd(mm, page, writable, referenced,
-				     none_or_zero, cr->result, unmapped);
+	trace_mm_scan_pmd(mm, page, writable, referenced, none_or_zero,
+			  cr->result, unmapped);
 }
 
 static void collect_mm_slot(struct mm_slot *mm_slot)
@@ -1431,7 +1428,7 @@ static void collect_mm_slot(struct mm_slot *mm_slot)
 
 	lockdep_assert_held(&khugepaged_mm_lock);
 
-	if (khugepaged_test_exit(mm)) {
+	if (test_exit(mm)) {
 		/* free mm_slot */
 		hash_del(&mm_slot->hash);
 		list_del(&mm_slot->mm_node);
@@ -1602,7 +1599,7 @@ static void khugepaged_collapse_pte_mapped_thps(struct mm_slot *mm_slot)
 	if (!mmap_write_trylock(mm))
 		return;
 
-	if (unlikely(khugepaged_test_exit(mm)))
+	if (unlikely(test_exit(mm)))
 		goto out;
 
 	for (i = 0; i < mm_slot->nr_pte_mapped_thp; i++)
@@ -1665,7 +1662,7 @@ static void retract_page_tables(struct address_space *mapping, pgoff_t pgoff)
 			 * it'll always mapped in small page size for uffd-wp
 			 * registered ranges.
 			 */
-			if (!khugepaged_test_exit(mm) && !userfaultfd_wp(vma))
+			if (!test_exit(mm) && !userfaultfd_wp(vma))
 				collapse_and_free_pmd(mm, vma, addr, pmd);
 			mmap_write_unlock(mm);
 		} else {
@@ -1723,7 +1720,7 @@ static void collapse_file(struct mm_struct *mm,
 
 	/* Only allocate from the target node */
 	gfp = alloc_hugepage_khugepaged_gfpmask() | __GFP_THISNODE;
-	node = khugepaged_find_target_node(cc);
+	node = find_target_node(cc);
 
 	new_page = cc->alloc_hpage(cc, gfp, node);
 	if (!new_page) {
@@ -2107,7 +2104,7 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 		}
 
 		node = page_to_nid(page);
-		if (khugepaged_scan_abort(node, cc)) {
+		if (scan_abort(node, cc)) {
 			cr->result = SCAN_SCAN_ABORT;
 			break;
 		}
@@ -2196,7 +2193,7 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
 	vma = NULL;
 	if (unlikely(!mmap_read_trylock(mm)))
 		goto breakouterloop_mmap_lock;
-	if (likely(!khugepaged_test_exit(mm)))
+	if (likely(!test_exit(mm)))
 		vma = find_vma(mm, khugepaged_scan.address);
 
 	progress++;
@@ -2204,7 +2201,7 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
 		unsigned long hstart, hend;
 
 		cond_resched();
-		if (unlikely(khugepaged_test_exit(mm))) {
+		if (unlikely(test_exit(mm))) {
 			progress++;
 			break;
 		}
@@ -2228,7 +2225,7 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
 		while (khugepaged_scan.address < hend) {
 			struct collapse_result cr = {0};
 			cond_resched();
-			if (unlikely(khugepaged_test_exit(mm)))
+			if (unlikely(test_exit(mm)))
 				goto breakouterloop;
 
 			VM_BUG_ON(khugepaged_scan.address < hstart ||
@@ -2244,9 +2241,8 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
 				khugepaged_scan_file(mm, file, pgoff, cc, &cr);
 				fput(file);
 			} else {
-				khugepaged_scan_pmd(mm, vma,
-						    khugepaged_scan.address,
-						    cc, &cr);
+				scan_pmd(mm, vma, khugepaged_scan.address, cc,
+					 &cr);
 			}
 			if (cr.result == SCAN_SUCCEED)
 				++khugepaged_pages_collapsed;
@@ -2270,7 +2266,7 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
 	 * Release the current mm_slot if this mm is about to die, or
 	 * if we scanned all vmas of this mm.
 	 */
-	if (khugepaged_test_exit(mm) || !vma) {
+	if (test_exit(mm) || !vma) {
 		/*
 		 * Make sure that if mm_users is reaching zero while
 		 * khugepaged runs here, khugepaged_exit will find
@@ -2541,11 +2537,11 @@ int madvise_collapse(struct vm_area_struct *vma, struct vm_area_struct **prev,
 		cond_resched();
 		memset(&cr, 0, sizeof(cr));
 
-		if (unlikely(khugepaged_test_exit(mm)))
+		if (unlikely(test_exit(mm)))
 			break;
 
 		memset(cc.node_load, 0, sizeof(cc.node_load));
-		khugepaged_scan_pmd(mm, vma, addr, &cc, &cr);
+		scan_pmd(mm, vma, addr, &cc, &cr);
 		if (cr.dropped_mmap_lock)
 			*prev = NULL;  /* tell madvise we dropped mmap_lock */
 
-- 
2.36.0.rc0.470.gd361397f0d-goog

[PATCH v2 07/12] mm/khugepaged: add flag to ignore khugepaged_max_ptes_*

[Zach O'Keefe]

Add enforce_pte_scan_limits flag to struct collapse_control that allows
context to ignore sysfs-controlled knobs:
khugepaged_max_ptes_[none|swap|shared].  Set this flag in khugepaged
collapse context to preserve existing khugepaged behavior and unset the
flag in madvise collapse context since the user presumably has reason to
believe the collapse will be beneficial.

Signed-off-by: Zach O'Keefe <zokeefe@google.com>
---
 mm/khugepaged.c | 32 ++++++++++++++++++++++----------
 1 file changed, 22 insertions(+), 10 deletions(-)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 716ba465b356..2f95f60431aa 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -87,6 +87,9 @@ static struct kmem_cache *mm_slot_cache __read_mostly;
 #define MAX_PTE_MAPPED_THP 8
 
 struct collapse_control {
+	/* Respect khugepaged_max_ptes_[none|swap|shared] */
+	bool enforce_pte_scan_limits;
+
 	/* Num pages scanned per node */
 	int node_load[MAX_NUMNODES];
 
@@ -631,6 +634,7 @@ static bool is_refcount_suitable(struct page *page)
 static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 					unsigned long address,
 					pte_t *pte,
+					struct collapse_control *cc,
 					struct list_head *compound_pagelist)
 {
 	struct page *page = NULL;
@@ -644,7 +648,8 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 		if (pte_none(pteval) || (pte_present(pteval) &&
 				is_zero_pfn(pte_pfn(pteval)))) {
 			if (!userfaultfd_armed(vma) &&
-			    ++none_or_zero <= khugepaged_max_ptes_none) {
+			    (++none_or_zero <= khugepaged_max_ptes_none ||
+			     !cc->enforce_pte_scan_limits)) {
 				continue;
 			} else {
 				result = SCAN_EXCEED_NONE_PTE;
@@ -664,8 +669,8 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 
 		VM_BUG_ON_PAGE(!PageAnon(page), page);
 
-		if (page_mapcount(page) > 1 &&
-				++shared > khugepaged_max_ptes_shared) {
+		if (cc->enforce_pte_scan_limits && page_mapcount(page) > 1 &&
+		    ++shared > khugepaged_max_ptes_shared) {
 			result = SCAN_EXCEED_SHARED_PTE;
 			count_vm_event(THP_SCAN_EXCEED_SHARED_PTE);
 			goto out;
@@ -1207,7 +1212,7 @@ static void collapse_huge_page(struct mm_struct *mm, unsigned long address,
 	mmu_notifier_invalidate_range_end(&range);
 
 	spin_lock(pte_ptl);
-	cr->result =  __collapse_huge_page_isolate(vma, address, pte,
+	cr->result =  __collapse_huge_page_isolate(vma, address, pte, cc,
 						   &compound_pagelist);
 	spin_unlock(pte_ptl);
 
@@ -1296,7 +1301,8 @@ static void scan_pmd(struct mm_struct *mm, struct vm_area_struct *vma,
 	     _pte++, _address += PAGE_SIZE) {
 		pte_t pteval = *_pte;
 		if (is_swap_pte(pteval)) {
-			if (++unmapped <= khugepaged_max_ptes_swap) {
+			if (++unmapped <= khugepaged_max_ptes_swap ||
+			    !cc->enforce_pte_scan_limits) {
 				/*
 				 * Always be strict with uffd-wp
 				 * enabled swap entries.  Please see
@@ -1315,7 +1321,8 @@ static void scan_pmd(struct mm_struct *mm, struct vm_area_struct *vma,
 		}
 		if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval))) {
 			if (!userfaultfd_armed(vma) &&
-			    ++none_or_zero <= khugepaged_max_ptes_none) {
+			    (++none_or_zero <= khugepaged_max_ptes_none ||
+			     !cc->enforce_pte_scan_limits)) {
 				continue;
 			} else {
 				cr->result = SCAN_EXCEED_NONE_PTE;
@@ -1345,8 +1352,9 @@ static void scan_pmd(struct mm_struct *mm, struct vm_area_struct *vma,
 			goto out_unmap;
 		}
 
-		if (page_mapcount(page) > 1 &&
-				++shared > khugepaged_max_ptes_shared) {
+		if (cc->enforce_pte_scan_limits &&
+		    page_mapcount(page) > 1 &&
+		    ++shared > khugepaged_max_ptes_shared) {
 			cr->result = SCAN_EXCEED_SHARED_PTE;
 			count_vm_event(THP_SCAN_EXCEED_SHARED_PTE);
 			goto out_unmap;
@@ -2086,7 +2094,8 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 			continue;
 
 		if (xa_is_value(page)) {
-			if (++swap > khugepaged_max_ptes_swap) {
+			if (cc->enforce_pte_scan_limits &&
+			    ++swap > khugepaged_max_ptes_swap) {
 				cr->result = SCAN_EXCEED_SWAP_PTE;
 				count_vm_event(THP_SCAN_EXCEED_SWAP_PTE);
 				break;
@@ -2137,7 +2146,8 @@ static void khugepaged_scan_file(struct mm_struct *mm,
 	rcu_read_unlock();
 
 	if (cr->result == SCAN_SUCCEED) {
-		if (present < HPAGE_PMD_NR - khugepaged_max_ptes_none) {
+		if (present < HPAGE_PMD_NR - khugepaged_max_ptes_none &&
+		    cc->enforce_pte_scan_limits) {
 			cr->result = SCAN_EXCEED_NONE_PTE;
 			count_vm_event(THP_SCAN_EXCEED_NONE_PTE);
 		} else {
@@ -2364,6 +2374,7 @@ static int khugepaged(void *none)
 {
 	struct mm_slot *mm_slot;
 	struct collapse_control cc = {
+		.enforce_pte_scan_limits = true,
 		.last_target_node = NUMA_NO_NODE,
 		.alloc_hpage = &khugepaged_alloc_page,
 	};
@@ -2505,6 +2516,7 @@ int madvise_collapse(struct vm_area_struct *vma, struct vm_area_struct **prev,
 		     unsigned long start, unsigned long end)
 {
 	struct collapse_control cc = {
+		.enforce_pte_scan_limits = false,
 		.last_target_node = NUMA_NO_NODE,
 		.hpage = NULL,
 		.alloc_hpage = &alloc_hpage,
-- 
2.36.0.rc0.470.gd361397f0d-goog

[PATCH v2 08/12] mm/khugepaged: add flag to ignore page young/referenced requirement

[Zach O'Keefe]

Add enforce_young flag to struct collapse_control that allows context to
ignore requirement that some pages in region being collapsed be young or
referenced.  Set this flag in khugepaged collapse context to preserve
existing khugepaged behavior and unset the flag in madvise collapse
context since the user presumably has reason to believe the collapse
will be beneficial.

Signed-off-by: Zach O'Keefe <zokeefe@google.com>
---
 mm/khugepaged.c | 24 ++++++++++++++++--------
 1 file changed, 16 insertions(+), 8 deletions(-)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 2f95f60431aa..b9bf15faba26 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -90,6 +90,9 @@ struct collapse_control {
 	/* Respect khugepaged_max_ptes_[none|swap|shared] */
 	bool enforce_pte_scan_limits;
 
+	/* Require memory to be young */
+	bool enforce_young;
+
 	/* Num pages scanned per node */
 	int node_load[MAX_NUMNODES];
 
@@ -737,9 +740,10 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 			list_add_tail(&page->lru, compound_pagelist);
 next:
 		/* There should be enough young pte to collapse the page */
-		if (pte_young(pteval) ||
-		    page_is_young(page) || PageReferenced(page) ||
-		    mmu_notifier_test_young(vma->vm_mm, address))
+		if (cc->enforce_young &&
+		    (pte_young(pteval) || page_is_young(page) ||
+		     PageReferenced(page) || mmu_notifier_test_young(vma->vm_mm,
+								     address)))
 			referenced++;
 
 		if (pte_write(pteval))
@@ -748,7 +752,7 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 
 	if (unlikely(!writable)) {
 		result = SCAN_PAGE_RO;
-	} else if (unlikely(!referenced)) {
+	} else if (unlikely(cc->enforce_young && !referenced)) {
 		result = SCAN_LACK_REFERENCED_PAGE;
 	} else {
 		result = SCAN_SUCCEED;
@@ -1408,14 +1412,16 @@ static void scan_pmd(struct mm_struct *mm, struct vm_area_struct *vma,
 			cr->result = SCAN_PAGE_COUNT;
 			goto out_unmap;
 		}
-		if (pte_young(pteval) ||
-		    page_is_young(page) || PageReferenced(page) ||
-		    mmu_notifier_test_young(vma->vm_mm, address))
+		if (cc->enforce_young &&
+		    (pte_young(pteval) || page_is_young(page) ||
+		     PageReferenced(page) || mmu_notifier_test_young(vma->vm_mm,
+								     address)))
 			referenced++;
 	}
 	if (!writable) {
 		cr->result = SCAN_PAGE_RO;
-	} else if (!referenced || (unmapped && referenced < HPAGE_PMD_NR/2)) {
+	} else if (cc->enforce_young && (!referenced || (unmapped && referenced
+							 < HPAGE_PMD_NR / 2))) {
 		cr->result = SCAN_LACK_REFERENCED_PAGE;
 	} else {
 		cr->result = SCAN_SUCCEED;
@@ -2375,6 +2381,7 @@ static int khugepaged(void *none)
 	struct mm_slot *mm_slot;
 	struct collapse_control cc = {
 		.enforce_pte_scan_limits = true,
+		.enforce_young = true,
 		.last_target_node = NUMA_NO_NODE,
 		.alloc_hpage = &khugepaged_alloc_page,
 	};
@@ -2517,6 +2524,7 @@ int madvise_collapse(struct vm_area_struct *vma, struct vm_area_struct **prev,
 {
 	struct collapse_control cc = {
 		.enforce_pte_scan_limits = false,
+		.enforce_young = false,
 		.last_target_node = NUMA_NO_NODE,
 		.hpage = NULL,
 		.alloc_hpage = &alloc_hpage,
-- 
2.36.0.rc0.470.gd361397f0d-goog

[PATCH v2 09/12] mm/madvise: add MADV_COLLAPSE to process_madvise()

[Zach O'Keefe]

Allow MADV_COLLAPSE behavior for process_madvise(2) if caller has
CAP_SYS_ADMIN or is requesting collapse of it's own memory.

Signed-off-by: Zach O'Keefe <zokeefe@google.com>
---
 mm/madvise.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/mm/madvise.c b/mm/madvise.c
index 7ad53e5311cf..a5c82fa7972b 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -1165,13 +1165,15 @@ madvise_behavior_valid(int behavior)
 }
 
 static bool
-process_madvise_behavior_valid(int behavior)
+process_madvise_behavior_valid(int behavior, struct task_struct *task)
 {
 	switch (behavior) {
 	case MADV_COLD:
 	case MADV_PAGEOUT:
 	case MADV_WILLNEED:
 		return true;
+	case MADV_COLLAPSE:
+		return task == current || capable(CAP_SYS_ADMIN);
 	default:
 		return false;
 	}
@@ -1449,7 +1451,7 @@ SYSCALL_DEFINE5(process_madvise, int, pidfd, const struct iovec __user *, vec,
 		goto free_iov;
 	}
 
-	if (!process_madvise_behavior_valid(behavior)) {
+	if (!process_madvise_behavior_valid(behavior, task)) {
 		ret = -EINVAL;
 		goto release_task;
 	}
-- 
2.36.0.rc0.470.gd361397f0d-goog

[PATCH v2 10/12] selftests/vm: modularize collapse selftests

[Zach O'Keefe]

Modularize the collapse action of khugepaged collapse selftests by
introducing a struct collapse_context which specifies how to collapse a
given memory range and the expected semantics of the collapse.  This
can be reused later to test other collapse contexts.

Signed-off-by: Zach O'Keefe <zokeefe@google.com>
---
 tools/testing/selftests/vm/khugepaged.c | 257 +++++++++++-------------
 1 file changed, 116 insertions(+), 141 deletions(-)

diff --git a/tools/testing/selftests/vm/khugepaged.c b/tools/testing/selftests/vm/khugepaged.c
index 155120b67a16..c59d832fee96 100644
--- a/tools/testing/selftests/vm/khugepaged.c
+++ b/tools/testing/selftests/vm/khugepaged.c
@@ -23,6 +23,12 @@ static int hpage_pmd_nr;
 #define THP_SYSFS "/sys/kernel/mm/transparent_hugepage/"
 #define PID_SMAPS "/proc/self/smaps"
 
+struct collapse_context {
+	const char *name;
+	void (*collapse)(const char *msg, char *p, bool expect);
+	bool enforce_pte_scan_limits;
+};
+
 enum thp_enabled {
 	THP_ALWAYS,
 	THP_MADVISE,
@@ -528,53 +534,39 @@ static void alloc_at_fault(void)
 	munmap(p, hpage_pmd_size);
 }
 
-static void collapse_full(void)
+static void collapse_full(struct collapse_context *context)
 {
 	void *p;
 
 	p = alloc_mapping();
 	fill_memory(p, 0, hpage_pmd_size);
-	if (wait_for_scan("Collapse fully populated PTE table", p))
-		fail("Timeout");
-	else if (check_huge(p))
-		success("OK");
-	else
-		fail("Fail");
+	context->collapse("Collapse fully populated PTE table", p, true);
 	validate_memory(p, 0, hpage_pmd_size);
 	munmap(p, hpage_pmd_size);
 }
 
-static void collapse_empty(void)
+static void collapse_empty(struct collapse_context *context)
 {
 	void *p;
 
 	p = alloc_mapping();
-	if (wait_for_scan("Do not collapse empty PTE table", p))
-		fail("Timeout");
-	else if (check_huge(p))
-		fail("Fail");
-	else
-		success("OK");
+	context->collapse("Do not collapse empty PTE table", p, false);
 	munmap(p, hpage_pmd_size);
 }
 
-static void collapse_single_pte_entry(void)
+static void collapse_single_pte_entry(struct collapse_context *context)
 {
 	void *p;
 
 	p = alloc_mapping();
 	fill_memory(p, 0, page_size);
-	if (wait_for_scan("Collapse PTE table with single PTE entry present", p))
-		fail("Timeout");
-	else if (check_huge(p))
-		success("OK");
-	else
-		fail("Fail");
+	context->collapse("Collapse PTE table with single PTE entry present", p,
+			  true);
 	validate_memory(p, 0, page_size);
 	munmap(p, hpage_pmd_size);
 }
 
-static void collapse_max_ptes_none(void)
+static void collapse_max_ptes_none(struct collapse_context *context)
 {
 	int max_ptes_none = hpage_pmd_nr / 2;
 	struct settings settings = default_settings;
@@ -586,28 +578,23 @@ static void collapse_max_ptes_none(void)
 	p = alloc_mapping();
 
 	fill_memory(p, 0, (hpage_pmd_nr - max_ptes_none - 1) * page_size);
-	if (wait_for_scan("Do not collapse with max_ptes_none exceeded", p))
-		fail("Timeout");
-	else if (check_huge(p))
-		fail("Fail");
-	else
-		success("OK");
+	context->collapse("Maybe collapse with max_ptes_none exceeded", p,
+			  !context->enforce_pte_scan_limits);
 	validate_memory(p, 0, (hpage_pmd_nr - max_ptes_none - 1) * page_size);
 
-	fill_memory(p, 0, (hpage_pmd_nr - max_ptes_none) * page_size);
-	if (wait_for_scan("Collapse with max_ptes_none PTEs empty", p))
-		fail("Timeout");
-	else if (check_huge(p))
-		success("OK");
-	else
-		fail("Fail");
-	validate_memory(p, 0, (hpage_pmd_nr - max_ptes_none) * page_size);
+	if (context->enforce_pte_scan_limits) {
+		fill_memory(p, 0, (hpage_pmd_nr - max_ptes_none) * page_size);
+		context->collapse("Collapse with max_ptes_none PTEs empty", p,
+				  true);
+		validate_memory(p, 0,
+				(hpage_pmd_nr - max_ptes_none) * page_size);
+	}
 
 	munmap(p, hpage_pmd_size);
 	write_settings(&default_settings);
 }
 
-static void collapse_swapin_single_pte(void)
+static void collapse_swapin_single_pte(struct collapse_context *context)
 {
 	void *p;
 	p = alloc_mapping();
@@ -625,18 +612,14 @@ static void collapse_swapin_single_pte(void)
 		goto out;
 	}
 
-	if (wait_for_scan("Collapse with swapping in single PTE entry", p))
-		fail("Timeout");
-	else if (check_huge(p))
-		success("OK");
-	else
-		fail("Fail");
+	context->collapse("Collapse with swapping in single PTE entry",
+			  p, true);
 	validate_memory(p, 0, hpage_pmd_size);
 out:
 	munmap(p, hpage_pmd_size);
 }
 
-static void collapse_max_ptes_swap(void)
+static void collapse_max_ptes_swap(struct collapse_context *context)
 {
 	int max_ptes_swap = read_num("khugepaged/max_ptes_swap");
 	void *p;
@@ -656,39 +639,34 @@ static void collapse_max_ptes_swap(void)
 		goto out;
 	}
 
-	if (wait_for_scan("Do not collapse with max_ptes_swap exceeded", p))
-		fail("Timeout");
-	else if (check_huge(p))
-		fail("Fail");
-	else
-		success("OK");
+	context->collapse("Maybe collapse with max_ptes_swap exceeded",
+			      p, !context->enforce_pte_scan_limits);
 	validate_memory(p, 0, hpage_pmd_size);
 
-	fill_memory(p, 0, hpage_pmd_size);
-	printf("Swapout %d of %d pages...", max_ptes_swap, hpage_pmd_nr);
-	if (madvise(p, max_ptes_swap * page_size, MADV_PAGEOUT)) {
-		perror("madvise(MADV_PAGEOUT)");
-		exit(EXIT_FAILURE);
-	}
-	if (check_swap(p, max_ptes_swap * page_size)) {
-		success("OK");
-	} else {
-		fail("Fail");
-		goto out;
-	}
+	if (context->enforce_pte_scan_limits) {
+		fill_memory(p, 0, hpage_pmd_size);
+		printf("Swapout %d of %d pages...", max_ptes_swap,
+		       hpage_pmd_nr);
+		if (madvise(p, max_ptes_swap * page_size, MADV_PAGEOUT)) {
+			perror("madvise(MADV_PAGEOUT)");
+			exit(EXIT_FAILURE);
+		}
+		if (check_swap(p, max_ptes_swap * page_size)) {
+			success("OK");
+		} else {
+			fail("Fail");
+			goto out;
+		}
 
-	if (wait_for_scan("Collapse with max_ptes_swap pages swapped out", p))
-		fail("Timeout");
-	else if (check_huge(p))
-		success("OK");
-	else
-		fail("Fail");
-	validate_memory(p, 0, hpage_pmd_size);
+		context->collapse("Collapse with max_ptes_swap pages swapped out",
+				  p, true);
+		validate_memory(p, 0, hpage_pmd_size);
+	}
 out:
 	munmap(p, hpage_pmd_size);
 }
 
-static void collapse_single_pte_entry_compound(void)
+static void collapse_single_pte_entry_compound(struct collapse_context *context)
 {
 	void *p;
 
@@ -710,17 +688,13 @@ static void collapse_single_pte_entry_compound(void)
 	else
 		fail("Fail");
 
-	if (wait_for_scan("Collapse PTE table with single PTE mapping compound page", p))
-		fail("Timeout");
-	else if (check_huge(p))
-		success("OK");
-	else
-		fail("Fail");
+	context->collapse("Collapse PTE table with single PTE mapping compound page",
+			  p, true);
 	validate_memory(p, 0, page_size);
 	munmap(p, hpage_pmd_size);
 }
 
-static void collapse_full_of_compound(void)
+static void collapse_full_of_compound(struct collapse_context *context)
 {
 	void *p;
 
@@ -742,17 +716,12 @@ static void collapse_full_of_compound(void)
 	else
 		fail("Fail");
 
-	if (wait_for_scan("Collapse PTE table full of compound pages", p))
-		fail("Timeout");
-	else if (check_huge(p))
-		success("OK");
-	else
-		fail("Fail");
+	context->collapse("Collapse PTE table full of compound pages", p, true);
 	validate_memory(p, 0, hpage_pmd_size);
 	munmap(p, hpage_pmd_size);
 }
 
-static void collapse_compound_extreme(void)
+static void collapse_compound_extreme(struct collapse_context *context)
 {
 	void *p;
 	int i;
@@ -798,18 +767,14 @@ static void collapse_compound_extreme(void)
 	else
 		fail("Fail");
 
-	if (wait_for_scan("Collapse PTE table full of different compound pages", p))
-		fail("Timeout");
-	else if (check_huge(p))
-		success("OK");
-	else
-		fail("Fail");
+	context->collapse("Collapse PTE table full of different compound pages",
+			  p, true);
 
 	validate_memory(p, 0, hpage_pmd_size);
 	munmap(p, hpage_pmd_size);
 }
 
-static void collapse_fork(void)
+static void collapse_fork(struct collapse_context *context)
 {
 	int wstatus;
 	void *p;
@@ -835,13 +800,8 @@ static void collapse_fork(void)
 			fail("Fail");
 
 		fill_memory(p, page_size, 2 * page_size);
-
-		if (wait_for_scan("Collapse PTE table with single page shared with parent process", p))
-			fail("Timeout");
-		else if (check_huge(p))
-			success("OK");
-		else
-			fail("Fail");
+		context->collapse("Collapse PTE table with single page shared with parent process",
+				  p, true);
 
 		validate_memory(p, 0, page_size);
 		munmap(p, hpage_pmd_size);
@@ -860,7 +820,7 @@ static void collapse_fork(void)
 	munmap(p, hpage_pmd_size);
 }
 
-static void collapse_fork_compound(void)
+static void collapse_fork_compound(struct collapse_context *context)
 {
 	int wstatus;
 	void *p;
@@ -896,14 +856,10 @@ static void collapse_fork_compound(void)
 		fill_memory(p, 0, page_size);
 
 		write_num("khugepaged/max_ptes_shared", hpage_pmd_nr - 1);
-		if (wait_for_scan("Collapse PTE table full of compound pages in child", p))
-			fail("Timeout");
-		else if (check_huge(p))
-			success("OK");
-		else
-			fail("Fail");
+		context->collapse("Collapse PTE table full of compound pages in child",
+				  p, true);
 		write_num("khugepaged/max_ptes_shared",
-				default_settings.khugepaged.max_ptes_shared);
+			  default_settings.khugepaged.max_ptes_shared);
 
 		validate_memory(p, 0, hpage_pmd_size);
 		munmap(p, hpage_pmd_size);
@@ -922,7 +878,7 @@ static void collapse_fork_compound(void)
 	munmap(p, hpage_pmd_size);
 }
 
-static void collapse_max_ptes_shared()
+static void collapse_max_ptes_shared(struct collapse_context *context)
 {
 	int max_ptes_shared = read_num("khugepaged/max_ptes_shared");
 	int wstatus;
@@ -957,28 +913,22 @@ static void collapse_max_ptes_shared()
 		else
 			fail("Fail");
 
-		if (wait_for_scan("Do not collapse with max_ptes_shared exceeded", p))
-			fail("Timeout");
-		else if (!check_huge(p))
-			success("OK");
-		else
-			fail("Fail");
-
-		printf("Trigger CoW on page %d of %d...",
-				hpage_pmd_nr - max_ptes_shared, hpage_pmd_nr);
-		fill_memory(p, 0, (hpage_pmd_nr - max_ptes_shared) * page_size);
-		if (!check_huge(p))
-			success("OK");
-		else
-			fail("Fail");
-
-
-		if (wait_for_scan("Collapse with max_ptes_shared PTEs shared", p))
-			fail("Timeout");
-		else if (check_huge(p))
-			success("OK");
-		else
-			fail("Fail");
+		context->collapse("Maybe collapse with max_ptes_shared exceeded",
+				  p, !context->enforce_pte_scan_limits);
+
+		if (context->enforce_pte_scan_limits) {
+			printf("Trigger CoW on page %d of %d...",
+			       hpage_pmd_nr - max_ptes_shared, hpage_pmd_nr);
+			fill_memory(p, 0, (hpage_pmd_nr - max_ptes_shared) *
+				    page_size);
+			if (!check_huge(p))
+				success("OK");
+			else
+				fail("Fail");
+
+			context->collapse("Collapse with max_ptes_shared PTEs shared",
+					  p, true);
+		}
 
 		validate_memory(p, 0, hpage_pmd_size);
 		munmap(p, hpage_pmd_size);
@@ -997,8 +947,27 @@ static void collapse_max_ptes_shared()
 	munmap(p, hpage_pmd_size);
 }
 
+static void khugepaged_collapse(const char *msg, char *p, bool expect)
+{
+	if (wait_for_scan(msg, p))
+		fail("Timeout");
+	else if (check_huge(p) == expect)
+		success("OK");
+	else
+		fail("Fail");
+}
+
 int main(void)
 {
+	struct collapse_context contexts[] = {
+		{
+			.name = "khugepaged",
+			.collapse = &khugepaged_collapse,
+			.enforce_pte_scan_limits = true,
+		},
+	};
+	int i;
+
 	setbuf(stdout, NULL);
 
 	page_size = getpagesize();
@@ -1014,18 +983,24 @@ int main(void)
 	adjust_settings();
 
 	alloc_at_fault();
-	collapse_full();
-	collapse_empty();
-	collapse_single_pte_entry();
-	collapse_max_ptes_none();
-	collapse_swapin_single_pte();
-	collapse_max_ptes_swap();
-	collapse_single_pte_entry_compound();
-	collapse_full_of_compound();
-	collapse_compound_extreme();
-	collapse_fork();
-	collapse_fork_compound();
-	collapse_max_ptes_shared();
+
+	for (i = 0; i < sizeof(contexts) / sizeof(contexts[0]); ++i) {
+		struct collapse_context *c = &contexts[i];
+
+		printf("\n*** Testing context: %s ***\n", c->name);
+		collapse_full(c);
+		collapse_empty(c);
+		collapse_single_pte_entry(c);
+		collapse_max_ptes_none(c);
+		collapse_swapin_single_pte(c);
+		collapse_max_ptes_swap(c);
+		collapse_single_pte_entry_compound(c);
+		collapse_full_of_compound(c);
+		collapse_compound_extreme(c);
+		collapse_fork(c);
+		collapse_fork_compound(c);
+		collapse_max_ptes_shared(c);
+	}
 
 	restore_settings(0);
 }
-- 
2.36.0.rc0.470.gd361397f0d-goog

[PATCH v2 11/12] selftests/vm: add MADV_COLLAPSE collapse context to selftests

[Zach O'Keefe]

Add MADV_COLLAPSE selftests.  Extend struct collapse_context to support
context initialization/cleanup.  This is used by madvise collapse
context to "disable" and "enable" khugepaged, since it would otherwise
interfere with the tests.

The mechanism used to "disable" khugepaged is a hack: it sets
/sys/kernel/mm/transparent_hugepage/khugepaged/scan_sleep_millisecs to a
large value and feeds khugepaged enough suitable VMAs/pages to keep
khugepaged sleeping for the duration of the madvise collapse tests.
Since khugepaged is woken when this file is written, enough VMAs must be
queued to put khugepaged back to sleep when the tests write to this file
in write_settings().

Signed-off-by: Zach O'Keefe <zokeefe@google.com>
---
 tools/testing/selftests/vm/khugepaged.c | 133 ++++++++++++++++++++++--
 1 file changed, 125 insertions(+), 8 deletions(-)

diff --git a/tools/testing/selftests/vm/khugepaged.c b/tools/testing/selftests/vm/khugepaged.c
index c59d832fee96..e0ccc9443f78 100644
--- a/tools/testing/selftests/vm/khugepaged.c
+++ b/tools/testing/selftests/vm/khugepaged.c
@@ -14,17 +14,23 @@
 #ifndef MADV_PAGEOUT
 #define MADV_PAGEOUT 21
 #endif
+#ifndef MADV_COLLAPSE
+#define MADV_COLLAPSE 25
+#endif
 
 #define BASE_ADDR ((void *)(1UL << 30))
 static unsigned long hpage_pmd_size;
 static unsigned long page_size;
 static int hpage_pmd_nr;
+static int num_khugepaged_wakeups;
 
 #define THP_SYSFS "/sys/kernel/mm/transparent_hugepage/"
 #define PID_SMAPS "/proc/self/smaps"
 
 struct collapse_context {
 	const char *name;
+	bool (*init_context)(void);
+	bool (*cleanup_context)(void);
 	void (*collapse)(const char *msg, char *p, bool expect);
 	bool enforce_pte_scan_limits;
 };
@@ -264,6 +270,17 @@ static void write_num(const char *name, unsigned long num)
 	}
 }
 
+/*
+ * Use this macro instead of write_settings inside tests, and should
+ * be called at most once per callsite.
+ *
+ * Hack to statically count the number of times khugepaged is woken up due to
+ * writes to
+ * /sys/kernel/mm/transparent_hugepage/khugepaged/scan_sleep_millisecs,
+ * and is stored in __COUNTER__.
+ */
+#define WRITE_SETTINGS(s) do { __COUNTER__; write_settings(s); } while (0)
+
 static void write_settings(struct settings *settings)
 {
 	struct khugepaged_settings *khugepaged = &settings->khugepaged;
@@ -332,7 +349,7 @@ static void adjust_settings(void)
 {
 
 	printf("Adjust settings...");
-	write_settings(&default_settings);
+	WRITE_SETTINGS(&default_settings);
 	success("OK");
 }
 
@@ -440,20 +457,25 @@ static bool check_swap(void *addr, unsigned long size)
 	return swap;
 }
 
-static void *alloc_mapping(void)
+static void *alloc_mapping_at(void *at, size_t size)
 {
 	void *p;
 
-	p = mmap(BASE_ADDR, hpage_pmd_size, PROT_READ | PROT_WRITE,
-			MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
-	if (p != BASE_ADDR) {
-		printf("Failed to allocate VMA at %p\n", BASE_ADDR);
+	p = mmap(at, size, PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_PRIVATE,
+		 -1, 0);
+	if (p != at) {
+		printf("Failed to allocate VMA at %p\n", at);
 		exit(EXIT_FAILURE);
 	}
 
 	return p;
 }
 
+static void *alloc_mapping(void)
+{
+	return alloc_mapping_at(BASE_ADDR, hpage_pmd_size);
+}
+
 static void fill_memory(int *p, unsigned long start, unsigned long end)
 {
 	int i;
@@ -573,7 +595,7 @@ static void collapse_max_ptes_none(struct collapse_context *context)
 	void *p;
 
 	settings.khugepaged.max_ptes_none = max_ptes_none;
-	write_settings(&settings);
+	WRITE_SETTINGS(&settings);
 
 	p = alloc_mapping();
 
@@ -591,7 +613,7 @@ static void collapse_max_ptes_none(struct collapse_context *context)
 	}
 
 	munmap(p, hpage_pmd_size);
-	write_settings(&default_settings);
+	WRITE_SETTINGS(&default_settings);
 }
 
 static void collapse_swapin_single_pte(struct collapse_context *context)
@@ -947,6 +969,87 @@ static void collapse_max_ptes_shared(struct collapse_context *context)
 	munmap(p, hpage_pmd_size);
 }
 
+static void madvise_collapse(const char *msg, char *p, bool expect)
+{
+	int ret;
+
+	printf("%s...", msg);
+	/* Sanity check */
+	if (check_huge(p)) {
+		printf("Unexpected huge page\n");
+		exit(EXIT_FAILURE);
+	}
+
+	madvise(p, hpage_pmd_size, MADV_HUGEPAGE);
+	ret = madvise(p, hpage_pmd_size, MADV_COLLAPSE);
+	if (((bool)ret) == expect)
+		fail("Fail: Bad return value");
+	else if (check_huge(p) != expect)
+		fail("Fail: check_huge()");
+	else
+		success("OK");
+}
+
+static struct khugepaged_disable_state {
+	void *p;
+	size_t map_size;
+} khugepaged_disable_state;
+
+static bool disable_khugepaged(void)
+{
+	/*
+	 * Hack to "disable" khugepaged by setting
+	 * /transparent_hugepage/khugepaged/scan_sleep_millisecs to some large
+	 * value, then feeding it enough suitable VMAs to scan and subsequently
+	 * sleep.
+	 *
+	 * khugepaged is woken up on writes to
+	 * /transparent_hugepage/khugepaged/scan_sleep_millisecs, so care must
+	 * be taken to not inadvertently wake khugepaged in these tests.
+	 *
+	 * Feed khugepaged 1 hugepage-sized VMA to scan and sleep on, then
+	 * N more for each time khugepaged would be woken up.
+	 */
+	size_t map_size = (num_khugepaged_wakeups + 1) * hpage_pmd_size;
+	void *p;
+	bool ret = true;
+	int full_scans;
+	int timeout = 6;  /* 3 seconds */
+
+	default_settings.khugepaged.scan_sleep_millisecs = 1000 * 60 * 10;
+	default_settings.khugepaged.pages_to_scan = 1;
+	write_settings(&default_settings);
+
+	p = alloc_mapping_at(((char *)BASE_ADDR) + (1UL << 30), map_size);
+	fill_memory(p, 0, map_size);
+
+	full_scans = read_num("khugepaged/full_scans") + 2;
+
+	printf("disabling khugepaged...");
+	while (timeout--) {
+		if (read_num("khugepaged/full_scans") >= full_scans) {
+			fail("Fail");
+			ret = false;
+			break;
+		}
+		printf(".");
+		usleep(TICK);
+	}
+	success("OK");
+	khugepaged_disable_state.p = p;
+	khugepaged_disable_state.map_size = map_size;
+	return ret;
+}
+
+static bool enable_khugepaged(void)
+{
+	printf("enabling khugepaged...");
+	munmap(khugepaged_disable_state.p, khugepaged_disable_state.map_size);
+	write_settings(&saved_settings);
+	success("OK");
+	return true;
+}
+
 static void khugepaged_collapse(const char *msg, char *p, bool expect)
 {
 	if (wait_for_scan(msg, p))
@@ -962,9 +1065,18 @@ int main(void)
 	struct collapse_context contexts[] = {
 		{
 			.name = "khugepaged",
+			.init_context = NULL,
+			.cleanup_context = NULL,
 			.collapse = &khugepaged_collapse,
 			.enforce_pte_scan_limits = true,
 		},
+		{
+			.name = "madvise",
+			.init_context = &disable_khugepaged,
+			.cleanup_context = &enable_khugepaged,
+			.collapse = &madvise_collapse,
+			.enforce_pte_scan_limits = false,
+		},
 	};
 	int i;
 
@@ -973,6 +1085,7 @@ int main(void)
 	page_size = getpagesize();
 	hpage_pmd_size = read_num("hpage_pmd_size");
 	hpage_pmd_nr = hpage_pmd_size / page_size;
+	num_khugepaged_wakeups = __COUNTER__;
 
 	default_settings.khugepaged.max_ptes_none = hpage_pmd_nr - 1;
 	default_settings.khugepaged.max_ptes_swap = hpage_pmd_nr / 8;
@@ -988,6 +1101,8 @@ int main(void)
 		struct collapse_context *c = &contexts[i];
 
 		printf("\n*** Testing context: %s ***\n", c->name);
+		if (c->init_context && !c->init_context())
+			continue;
 		collapse_full(c);
 		collapse_empty(c);
 		collapse_single_pte_entry(c);
@@ -1000,6 +1115,8 @@ int main(void)
 		collapse_fork(c);
 		collapse_fork_compound(c);
 		collapse_max_ptes_shared(c);
+		if (c->cleanup_context && !c->cleanup_context())
+			break;
 	}
 
 	restore_settings(0);
-- 
2.36.0.rc0.470.gd361397f0d-goog

[PATCH v2 12/12] selftests/vm: add test to verify recollapse of THPs

[Zach O'Keefe]

Add selftest specific to madvise collapse context that tests
MADV_COLLAPSE is "successful" if a hugepage-algined/sized region is
already pmd-mapped.

Signed-off-by: Zach O'Keefe <zokeefe@google.com>
---
 tools/testing/selftests/vm/khugepaged.c | 32 +++++++++++++++++++++++++
 1 file changed, 32 insertions(+)

diff --git a/tools/testing/selftests/vm/khugepaged.c b/tools/testing/selftests/vm/khugepaged.c
index e0ccc9443f78..c36d04218083 100644
--- a/tools/testing/selftests/vm/khugepaged.c
+++ b/tools/testing/selftests/vm/khugepaged.c
@@ -969,6 +969,32 @@ static void collapse_max_ptes_shared(struct collapse_context *context)
 	munmap(p, hpage_pmd_size);
 }
 
+static void madvise_collapse_existing_thps(void)
+{
+	void *p;
+	int err;
+
+	p = alloc_mapping();
+	fill_memory(p, 0, hpage_pmd_size);
+
+	printf("Collapse fully populated PTE table...");
+	madvise(p, hpage_pmd_size, MADV_HUGEPAGE);
+	err = madvise(p, hpage_pmd_size, MADV_COLLAPSE);
+	if (err == 0 && check_huge(p)) {
+		success("OK");
+		printf("Re-collapse PMD-mapped hugepage");
+		err = madvise(p, hpage_pmd_size, MADV_COLLAPSE);
+		if (err == 0 && check_huge(p))
+			success("OK");
+		else
+			fail("Fail");
+	} else {
+		fail("Fail");
+	}
+	validate_memory(p, 0, hpage_pmd_size);
+	munmap(p, hpage_pmd_size);
+}
+
 static void madvise_collapse(const char *msg, char *p, bool expect)
 {
 	int ret;
@@ -1097,6 +1123,7 @@ int main(void)
 
 	alloc_at_fault();
 
+	/* Shared tests */
 	for (i = 0; i < sizeof(contexts) / sizeof(contexts[0]); ++i) {
 		struct collapse_context *c = &contexts[i];
 
@@ -1119,5 +1146,10 @@ int main(void)
 			break;
 	}
 
+	/* madvise-specific tests */
+	disable_khugepaged();
+	madvise_collapse_existing_thps();
+	enable_khugepaged();
+
 	restore_settings(0);
 }
-- 
2.36.0.rc0.470.gd361397f0d-goog

Re: [PATCH v2 00/12] mm: userspace hugepage collapse

[Peter Xu]

Hi, Zach,

On Thu, Apr 14, 2022 at 11:06:00AM -0700, Zach O'Keefe wrote:
> process_madvise(2)
> 
> 	Performs a synchronous collapse of the native pages
> 	mapped by the list of iovecs into transparent hugepages.
> 
> 	Allocation semantics are the same as khugepaged, and depend on
> 	(1) the active sysfs settings
> 	/sys/kernel/mm/transparent_hugepage/enabled and
> 	/sys/kernel/mm/transparent_hugepage/khugepaged/defrag, and (2)
> 	the VMA flags of the memory range being collapsed.
> 
> 	Collapse eligibility criteria differs from khugepaged in that
> 	the sysfs files
> 	/sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_[none|swap|shared]
> 	are ignored.

The userspace khugepaged idea definitely makes sense to me, though I'm
curious how the line is drown on the different behaviors here by explicitly
ignoring the max_ptes_* entries.

Let's assume the initiative is to duplicate a more data-aware khugepaged in
the userspace, then IMHO it makes more sense to start with all the policies
that applies to khugepaged already, including max_pte_*.

I can understand the willingness to provide even stronger semantics here
than khugepaged since the userspace could have very clear knowledge of how
to provision the memories (better than a kernel scanner).  It's just that
IMHO it could be slightly confusing if the new interface only partially
apply the khugepaged rules.

No strong opinion here.  It could already been a trade-off after the
discussion from the RFC with Michal which I read..  Just curious about how
you made that design decision so feel free to read it as a pure question.

Thanks,

-- 
Peter Xu

Re: [PATCH v2 00/12] mm: userspace hugepage collapse

[Zach O'Keefe]

Hey Peter,

Thanks for taking the time to review!

On Thu, Apr 14, 2022 at 5:04 PM Peter Xu <peterx@redhat.com> wrote:
>
> Hi, Zach,
>
> On Thu, Apr 14, 2022 at 11:06:00AM -0700, Zach O'Keefe wrote:
> > process_madvise(2)
> >
> >       Performs a synchronous collapse of the native pages
> >       mapped by the list of iovecs into transparent hugepages.
> >
> >       Allocation semantics are the same as khugepaged, and depend on
> >       (1) the active sysfs settings
> >       /sys/kernel/mm/transparent_hugepage/enabled and
> >       /sys/kernel/mm/transparent_hugepage/khugepaged/defrag, and (2)
> >       the VMA flags of the memory range being collapsed.
> >
> >       Collapse eligibility criteria differs from khugepaged in that
> >       the sysfs files
> >       /sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_[none|swap|shared]
> >       are ignored.
>
> The userspace khugepaged idea definitely makes sense to me, though I'm
> curious how the line is drown on the different behaviors here by explicitly
> ignoring the max_ptes_* entries.
>
> Let's assume the initiative is to duplicate a more data-aware khugepaged in
> the userspace, then IMHO it makes more sense to start with all the policies
> that applies to khugepaged already, including max_pte_*.
>
> I can understand the willingness to provide even stronger semantics here
> than khugepaged since the userspace could have very clear knowledge of how
> to provision the memories (better than a kernel scanner).  It's just that
> IMHO it could be slightly confusing if the new interface only partially
> apply the khugepaged rules.
>
> No strong opinion here.  It could already been a trade-off after the
> discussion from the RFC with Michal which I read..  Just curious about how
> you made that design decision so feel free to read it as a pure question.
>

Understand your point here. The allocation and max_pte_* semantics are
split between khugepaged-like and fault-like, respectively - which
could be confusing. Originally, I proposed a MADV_F_COLLAPSE_LIMITS
flag to control the former's behavior, but agreed to keep things
simple to start, and expand the interface if/when necessary. I opted
to ignore max_ptes_* as the default since I envisioned that early
adopters would "just want it to work". One such example would be
backing executable text by hugepages on program load when many pages
haven't been demand-paged in yet.

What do you think?

Thanks,
Zach

> Thanks,
>
> --
> Peter Xu
>

Re: [PATCH v2 00/12] mm: userspace hugepage collapse

[Peter Xu]

On Thu, Apr 14, 2022 at 05:52:43PM -0700, Zach O'Keefe wrote:
> Hey Peter,
> 
> Thanks for taking the time to review!
> 
> On Thu, Apr 14, 2022 at 5:04 PM Peter Xu <peterx@redhat.com> wrote:
> >
> > Hi, Zach,
> >
> > On Thu, Apr 14, 2022 at 11:06:00AM -0700, Zach O'Keefe wrote:
> > > process_madvise(2)
> > >
> > >       Performs a synchronous collapse of the native pages
> > >       mapped by the list of iovecs into transparent hugepages.
> > >
> > >       Allocation semantics are the same as khugepaged, and depend on
> > >       (1) the active sysfs settings
> > >       /sys/kernel/mm/transparent_hugepage/enabled and
> > >       /sys/kernel/mm/transparent_hugepage/khugepaged/defrag, and (2)
> > >       the VMA flags of the memory range being collapsed.
> > >
> > >       Collapse eligibility criteria differs from khugepaged in that
> > >       the sysfs files
> > >       /sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_[none|swap|shared]
> > >       are ignored.
> >
> > The userspace khugepaged idea definitely makes sense to me, though I'm
> > curious how the line is drown on the different behaviors here by explicitly
> > ignoring the max_ptes_* entries.
> >
> > Let's assume the initiative is to duplicate a more data-aware khugepaged in
> > the userspace, then IMHO it makes more sense to start with all the policies
> > that applies to khugepaged already, including max_pte_*.
> >
> > I can understand the willingness to provide even stronger semantics here
> > than khugepaged since the userspace could have very clear knowledge of how
> > to provision the memories (better than a kernel scanner).  It's just that
> > IMHO it could be slightly confusing if the new interface only partially
> > apply the khugepaged rules.
> >
> > No strong opinion here.  It could already been a trade-off after the
> > discussion from the RFC with Michal which I read..  Just curious about how
> > you made that design decision so feel free to read it as a pure question.
> >
> 
> Understand your point here. The allocation and max_pte_* semantics are
> split between khugepaged-like and fault-like, respectively - which
> could be confusing. Originally, I proposed a MADV_F_COLLAPSE_LIMITS
> flag to control the former's behavior, but agreed to keep things
> simple to start, and expand the interface if/when necessary. I opted
> to ignore max_ptes_* as the default since I envisioned that early
> adopters would "just want it to work". One such example would be
> backing executable text by hugepages on program load when many pages
> haven't been demand-paged in yet.
> 
> What do you think?

I'm just slightly worried that'll make the default MADV_COLLAPSE semantics
blurred.

To me, a clean default definition for MADV_COLLAPSE would be nice, as "do
khugepaged on this range, and with current thread context".  IMHO any
feature bits then can be supplementing special needs, and I'll take the thp
backing executable example to be one of the (good?) reason we'd need an
extra flag for ignoring the max_ptes_* knobs.

So personally if I were you maybe I'll start with the simple scheme of that
(even if it won't immediately service a thing) but then add either the
defrag or ignore_max_ptes_* as feature bits later on, with clear use case
descriptions about why we need each of the feature flags.  IMHO numbers
would be even more helpful when there's specific use cases on the show.

Or, perhaps you think all potential MADV_COLLAPSE users should literally
skip max_ptes_* limitations always?

Anyway, I won't pretend I am an expert in this area. :) So please take that
with a grain of salt.

Thanks,

-- 
Peter Xu

Re: [PATCH v2 00/12] mm: userspace hugepage collapse

[Zach O'Keefe]

On Fri, Apr 15, 2022 at 6:39 AM Peter Xu <peterx@redhat.com> wrote:
>
> On Thu, Apr 14, 2022 at 05:52:43PM -0700, Zach O'Keefe wrote:
> > Hey Peter,
> >
> > Thanks for taking the time to review!
> >
> > On Thu, Apr 14, 2022 at 5:04 PM Peter Xu <peterx@redhat.com> wrote:
> > >
> > > Hi, Zach,
> > >
> > > On Thu, Apr 14, 2022 at 11:06:00AM -0700, Zach O'Keefe wrote:
> > > > process_madvise(2)
> > > >
> > > >       Performs a synchronous collapse of the native pages
> > > >       mapped by the list of iovecs into transparent hugepages.
> > > >
> > > >       Allocation semantics are the same as khugepaged, and depend on
> > > >       (1) the active sysfs settings
> > > >       /sys/kernel/mm/transparent_hugepage/enabled and
> > > >       /sys/kernel/mm/transparent_hugepage/khugepaged/defrag, and (2)
> > > >       the VMA flags of the memory range being collapsed.
> > > >
> > > >       Collapse eligibility criteria differs from khugepaged in that
> > > >       the sysfs files
> > > >       /sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_[none|swap|shared]
> > > >       are ignored.
> > >
> > > The userspace khugepaged idea definitely makes sense to me, though I'm
> > > curious how the line is drown on the different behaviors here by explicitly
> > > ignoring the max_ptes_* entries.
> > >
> > > Let's assume the initiative is to duplicate a more data-aware khugepaged in
> > > the userspace, then IMHO it makes more sense to start with all the policies
> > > that applies to khugepaged already, including max_pte_*.
> > >
> > > I can understand the willingness to provide even stronger semantics here
> > > than khugepaged since the userspace could have very clear knowledge of how
> > > to provision the memories (better than a kernel scanner).  It's just that
> > > IMHO it could be slightly confusing if the new interface only partially
> > > apply the khugepaged rules.
> > >
> > > No strong opinion here.  It could already been a trade-off after the
> > > discussion from the RFC with Michal which I read..  Just curious about how
> > > you made that design decision so feel free to read it as a pure question.
> > >
> >
> > Understand your point here. The allocation and max_pte_* semantics are
> > split between khugepaged-like and fault-like, respectively - which
> > could be confusing. Originally, I proposed a MADV_F_COLLAPSE_LIMITS
> > flag to control the former's behavior, but agreed to keep things
> > simple to start, and expand the interface if/when necessary. I opted
> > to ignore max_ptes_* as the default since I envisioned that early
> > adopters would "just want it to work". One such example would be
> > backing executable text by hugepages on program load when many pages
> > haven't been demand-paged in yet.
> >
> > What do you think?
>
> I'm just slightly worried that'll make the default MADV_COLLAPSE semantics
> blurred.
>
> To me, a clean default definition for MADV_COLLAPSE would be nice, as "do
> khugepaged on this range, and with current thread context".  IMHO any
> feature bits then can be supplementing special needs, and I'll take the thp
> backing executable example to be one of the (good?) reason we'd need an
> extra flag for ignoring the max_ptes_* knobs.
>
> So personally if I were you maybe I'll start with the simple scheme of that
> (even if it won't immediately service a thing) but then add either the
> defrag or ignore_max_ptes_* as feature bits later on, with clear use case
> descriptions about why we need each of the feature flags.  IMHO numbers
> would be even more helpful when there's specific use cases on the show.
>
> Or, perhaps you think all potential MADV_COLLAPSE users should literally
> skip max_ptes_* limitations always?
>

Thanks for your time and valuable feedback here, Peter. I had a response typed
up, but after a few iterations became increasingly unsatisfied with my
own response.

I think this feature should be able to stand on its own without
consideration of a userspace khugepaged, as we have existing concrete
examples where it would be useful. In these cases, and I assume almost
all other use-cases outside userspace khugepaged, max_ptes_* should be
ignored as the fundamental assumption of MADV_COLLAPSE is that the
user knows better, and IMHO, khugepaged heuristics shouldn't tell
users they are wrong.

But this, as you mention, unsatisfactorily blurs the semantics of
MADV_COLLAPSE: "act like khugepaged here, but not here".

As such, WDYT about the reverse-side of the coin of what you proposed:
to not couple the default behavior of MADV_COLLAPSE with khugepaged at
all? I.e. Not tie the allocation semantics to
/sys/kernel/mm/transparent_hugepage/khugepaged/defrag. We can add
flags as necessary when/if a reimplementation of khugepaged in
userspace proves fruitful.

Thanks for your time and input,
Zach



> Anyway, I won't pretend I am an expert in this area. :) So please take that
> with a grain of salt.
>
> Thanks,
>
> --
> Peter Xu
>

Re: [PATCH v2 00/12] mm: userspace hugepage collapse

[Peter Xu]

Hi, Zach,

On Fri, Apr 15, 2022 at 01:04:04PM -0700, Zach O'Keefe wrote:
> On Fri, Apr 15, 2022 at 6:39 AM Peter Xu <peterx@redhat.com> wrote:
> >
> > On Thu, Apr 14, 2022 at 05:52:43PM -0700, Zach O'Keefe wrote:
> > > Hey Peter,
> > >
> > > Thanks for taking the time to review!
> > >
> > > On Thu, Apr 14, 2022 at 5:04 PM Peter Xu <peterx@redhat.com> wrote:
> > > >
> > > > Hi, Zach,
> > > >
> > > > On Thu, Apr 14, 2022 at 11:06:00AM -0700, Zach O'Keefe wrote:
> > > > > process_madvise(2)
> > > > >
> > > > >       Performs a synchronous collapse of the native pages
> > > > >       mapped by the list of iovecs into transparent hugepages.
> > > > >
> > > > >       Allocation semantics are the same as khugepaged, and depend on
> > > > >       (1) the active sysfs settings
> > > > >       /sys/kernel/mm/transparent_hugepage/enabled and
> > > > >       /sys/kernel/mm/transparent_hugepage/khugepaged/defrag, and (2)
> > > > >       the VMA flags of the memory range being collapsed.
> > > > >
> > > > >       Collapse eligibility criteria differs from khugepaged in that
> > > > >       the sysfs files
> > > > >       /sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_[none|swap|shared]
> > > > >       are ignored.
> > > >
> > > > The userspace khugepaged idea definitely makes sense to me, though I'm
> > > > curious how the line is drown on the different behaviors here by explicitly
> > > > ignoring the max_ptes_* entries.
> > > >
> > > > Let's assume the initiative is to duplicate a more data-aware khugepaged in
> > > > the userspace, then IMHO it makes more sense to start with all the policies
> > > > that applies to khugepaged already, including max_pte_*.
> > > >
> > > > I can understand the willingness to provide even stronger semantics here
> > > > than khugepaged since the userspace could have very clear knowledge of how
> > > > to provision the memories (better than a kernel scanner).  It's just that
> > > > IMHO it could be slightly confusing if the new interface only partially
> > > > apply the khugepaged rules.
> > > >
> > > > No strong opinion here.  It could already been a trade-off after the
> > > > discussion from the RFC with Michal which I read..  Just curious about how
> > > > you made that design decision so feel free to read it as a pure question.
> > > >
> > >
> > > Understand your point here. The allocation and max_pte_* semantics are
> > > split between khugepaged-like and fault-like, respectively - which
> > > could be confusing. Originally, I proposed a MADV_F_COLLAPSE_LIMITS
> > > flag to control the former's behavior, but agreed to keep things
> > > simple to start, and expand the interface if/when necessary. I opted
> > > to ignore max_ptes_* as the default since I envisioned that early
> > > adopters would "just want it to work". One such example would be
> > > backing executable text by hugepages on program load when many pages
> > > haven't been demand-paged in yet.
> > >
> > > What do you think?
> >
> > I'm just slightly worried that'll make the default MADV_COLLAPSE semantics
> > blurred.
> >
> > To me, a clean default definition for MADV_COLLAPSE would be nice, as "do
> > khugepaged on this range, and with current thread context".  IMHO any
> > feature bits then can be supplementing special needs, and I'll take the thp
> > backing executable example to be one of the (good?) reason we'd need an
> > extra flag for ignoring the max_ptes_* knobs.
> >
> > So personally if I were you maybe I'll start with the simple scheme of that
> > (even if it won't immediately service a thing) but then add either the
> > defrag or ignore_max_ptes_* as feature bits later on, with clear use case
> > descriptions about why we need each of the feature flags.  IMHO numbers
> > would be even more helpful when there's specific use cases on the show.
> >
> > Or, perhaps you think all potential MADV_COLLAPSE users should literally
> > skip max_ptes_* limitations always?
> >
> 
> Thanks for your time and valuable feedback here, Peter. I had a response typed
> up, but after a few iterations became increasingly unsatisfied with my
> own response.
> 
> I think this feature should be able to stand on its own without
> consideration of a userspace khugepaged, as we have existing concrete
> examples where it would be useful. In these cases, and I assume almost
> all other use-cases outside userspace khugepaged, max_ptes_* should be
> ignored as the fundamental assumption of MADV_COLLAPSE is that the
> user knows better, and IMHO, khugepaged heuristics shouldn't tell
> users they are wrong.

Valid point.  And actually right after I replied I thought similarly on
whether we need to connect the two interfaces at all..

It's just that it's very easy to go think like that after reading the cover
letter since that's exactly what it is comparing to. :)

There's definitely a difference view on user/kernel level of things, then
it sounds reasonable to me if we add a new interface it by default has a
stronger semantics otherwise we may not bother if with MADV_HUGEPAGE's
existance.

So maybe max_ptes_* won't even make sense for MADV_COLLAPSE in most cases
as you said.  And that's a real pure question I asked above, and I feel
like your answer is actually "yes" we should always ignore the max_ptes_*
fields until there's a proof that it'll be helpful.

> 
> But this, as you mention, unsatisfactorily blurs the semantics of
> MADV_COLLAPSE: "act like khugepaged here, but not here".
> 
> As such, WDYT about the reverse-side of the coin of what you proposed:
> to not couple the default behavior of MADV_COLLAPSE with khugepaged at
> all? I.e. Not tie the allocation semantics to
> /sys/kernel/mm/transparent_hugepage/khugepaged/defrag. We can add
> flags as necessary when/if a reimplementation of khugepaged in
> userspace proves fruitful.

Let's see whether others have thoughts, but what you proposed here makes
sense to me.

Thanks,

-- 
Peter Xu

Re: [PATCH v2 00/12] mm: userspace hugepage collapse

[Zach O'Keefe]

On Sat, Apr 16, 2022 at 12:26 PM Peter Xu <peterx@redhat.com> wrote:
>
> Hi, Zach,
>
> On Fri, Apr 15, 2022 at 01:04:04PM -0700, Zach O'Keefe wrote:
> > On Fri, Apr 15, 2022 at 6:39 AM Peter Xu <peterx@redhat.com> wrote:
> > >
> > > On Thu, Apr 14, 2022 at 05:52:43PM -0700, Zach O'Keefe wrote:
> > > > Hey Peter,
> > > >
> > > > Thanks for taking the time to review!
> > > >
> > > > On Thu, Apr 14, 2022 at 5:04 PM Peter Xu <peterx@redhat.com> wrote:
> > > > >
> > > > > Hi, Zach,
> > > > >
> > > > > On Thu, Apr 14, 2022 at 11:06:00AM -0700, Zach O'Keefe wrote:
> > > > > > process_madvise(2)
> > > > > >
> > > > > >       Performs a synchronous collapse of the native pages
> > > > > >       mapped by the list of iovecs into transparent hugepages.
> > > > > >
> > > > > >       Allocation semantics are the same as khugepaged, and depend on
> > > > > >       (1) the active sysfs settings
> > > > > >       /sys/kernel/mm/transparent_hugepage/enabled and
> > > > > >       /sys/kernel/mm/transparent_hugepage/khugepaged/defrag, and (2)
> > > > > >       the VMA flags of the memory range being collapsed.
> > > > > >
> > > > > >       Collapse eligibility criteria differs from khugepaged in that
> > > > > >       the sysfs files
> > > > > >       /sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_[none|swap|shared]
> > > > > >       are ignored.
> > > > >
> > > > > The userspace khugepaged idea definitely makes sense to me, though I'm
> > > > > curious how the line is drown on the different behaviors here by explicitly
> > > > > ignoring the max_ptes_* entries.
> > > > >
> > > > > Let's assume the initiative is to duplicate a more data-aware khugepaged in
> > > > > the userspace, then IMHO it makes more sense to start with all the policies
> > > > > that applies to khugepaged already, including max_pte_*.
> > > > >
> > > > > I can understand the willingness to provide even stronger semantics here
> > > > > than khugepaged since the userspace could have very clear knowledge of how
> > > > > to provision the memories (better than a kernel scanner).  It's just that
> > > > > IMHO it could be slightly confusing if the new interface only partially
> > > > > apply the khugepaged rules.
> > > > >
> > > > > No strong opinion here.  It could already been a trade-off after the
> > > > > discussion from the RFC with Michal which I read..  Just curious about how
> > > > > you made that design decision so feel free to read it as a pure question.
> > > > >
> > > >
> > > > Understand your point here. The allocation and max_pte_* semantics are
> > > > split between khugepaged-like and fault-like, respectively - which
> > > > could be confusing. Originally, I proposed a MADV_F_COLLAPSE_LIMITS
> > > > flag to control the former's behavior, but agreed to keep things
> > > > simple to start, and expand the interface if/when necessary. I opted
> > > > to ignore max_ptes_* as the default since I envisioned that early
> > > > adopters would "just want it to work". One such example would be
> > > > backing executable text by hugepages on program load when many pages
> > > > haven't been demand-paged in yet.
> > > >
> > > > What do you think?
> > >
> > > I'm just slightly worried that'll make the default MADV_COLLAPSE semantics
> > > blurred.
> > >
> > > To me, a clean default definition for MADV_COLLAPSE would be nice, as "do
> > > khugepaged on this range, and with current thread context".  IMHO any
> > > feature bits then can be supplementing special needs, and I'll take the thp
> > > backing executable example to be one of the (good?) reason we'd need an
> > > extra flag for ignoring the max_ptes_* knobs.
> > >
> > > So personally if I were you maybe I'll start with the simple scheme of that
> > > (even if it won't immediately service a thing) but then add either the
> > > defrag or ignore_max_ptes_* as feature bits later on, with clear use case
> > > descriptions about why we need each of the feature flags.  IMHO numbers
> > > would be even more helpful when there's specific use cases on the show.
> > >
> > > Or, perhaps you think all potential MADV_COLLAPSE users should literally
> > > skip max_ptes_* limitations always?
> > >
> >
> > Thanks for your time and valuable feedback here, Peter. I had a response typed
> > up, but after a few iterations became increasingly unsatisfied with my
> > own response.
> >
> > I think this feature should be able to stand on its own without
> > consideration of a userspace khugepaged, as we have existing concrete
> > examples where it would be useful. In these cases, and I assume almost
> > all other use-cases outside userspace khugepaged, max_ptes_* should be
> > ignored as the fundamental assumption of MADV_COLLAPSE is that the
> > user knows better, and IMHO, khugepaged heuristics shouldn't tell
> > users they are wrong.
>
> Valid point.  And actually right after I replied I thought similarly on
> whether we need to connect the two interfaces at all..
>
> It's just that it's very easy to go think like that after reading the cover
> letter since that's exactly what it is comparing to. :)
>

Yes, this is my fault :) After others have had a chance to review /
align, I'll include the immediate use cases
in the v3 cover letter as well, rather than deep in individual patch
messages.

> There's definitely a difference view on user/kernel level of things, then
> it sounds reasonable to me if we add a new interface it by default has a
> stronger semantics otherwise we may not bother if with MADV_HUGEPAGE's
> existance.
>

Yes, good point.

> So maybe max_ptes_* won't even make sense for MADV_COLLAPSE in most cases
> as you said.  And that's a real pure question I asked above, and I feel
> like your answer is actually "yes" we should always ignore the max_ptes_*
> fields until there's a proof that it'll be helpful.
>
> >
> > But this, as you mention, unsatisfactorily blurs the semantics of
> > MADV_COLLAPSE: "act like khugepaged here, but not here".
> >
> > As such, WDYT about the reverse-side of the coin of what you proposed:
> > to not couple the default behavior of MADV_COLLAPSE with khugepaged at
> > all? I.e. Not tie the allocation semantics to
> > /sys/kernel/mm/transparent_hugepage/khugepaged/defrag. We can add
> > flags as necessary when/if a reimplementation of khugepaged in
> > userspace proves fruitful.
>
> Let's see whether others have thoughts, but what you proposed here makes
> sense to me.
>

Great! Sounds good to me. Thank you again for your time, questions,
and feedback!

Best,

Zach

> Thanks,
>
> --
> Peter Xu
>

Re: [PATCH v2 00/12] mm: userspace hugepage collapse

[David Hildenbrand]

On 16.04.22 21:26, Peter Xu wrote:
> Hi, Zach,
> 
> On Fri, Apr 15, 2022 at 01:04:04PM -0700, Zach O'Keefe wrote:
>> On Fri, Apr 15, 2022 at 6:39 AM Peter Xu <peterx@redhat.com> wrote:
>>>
>>> On Thu, Apr 14, 2022 at 05:52:43PM -0700, Zach O'Keefe wrote:
>>>> Hey Peter,
>>>>
>>>> Thanks for taking the time to review!
>>>>
>>>> On Thu, Apr 14, 2022 at 5:04 PM Peter Xu <peterx@redhat.com> wrote:
>>>>>
>>>>> Hi, Zach,
>>>>>
>>>>> On Thu, Apr 14, 2022 at 11:06:00AM -0700, Zach O'Keefe wrote:
>>>>>> process_madvise(2)
>>>>>>
>>>>>>       Performs a synchronous collapse of the native pages
>>>>>>       mapped by the list of iovecs into transparent hugepages.
>>>>>>
>>>>>>       Allocation semantics are the same as khugepaged, and depend on
>>>>>>       (1) the active sysfs settings
>>>>>>       /sys/kernel/mm/transparent_hugepage/enabled and
>>>>>>       /sys/kernel/mm/transparent_hugepage/khugepaged/defrag, and (2)
>>>>>>       the VMA flags of the memory range being collapsed.
>>>>>>
>>>>>>       Collapse eligibility criteria differs from khugepaged in that
>>>>>>       the sysfs files
>>>>>>       /sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_[none|swap|shared]
>>>>>>       are ignored.
>>>>>
>>>>> The userspace khugepaged idea definitely makes sense to me, though I'm
>>>>> curious how the line is drown on the different behaviors here by explicitly
>>>>> ignoring the max_ptes_* entries.
>>>>>
>>>>> Let's assume the initiative is to duplicate a more data-aware khugepaged in
>>>>> the userspace, then IMHO it makes more sense to start with all the policies
>>>>> that applies to khugepaged already, including max_pte_*.
>>>>>
>>>>> I can understand the willingness to provide even stronger semantics here
>>>>> than khugepaged since the userspace could have very clear knowledge of how
>>>>> to provision the memories (better than a kernel scanner).  It's just that
>>>>> IMHO it could be slightly confusing if the new interface only partially
>>>>> apply the khugepaged rules.
>>>>>
>>>>> No strong opinion here.  It could already been a trade-off after the
>>>>> discussion from the RFC with Michal which I read..  Just curious about how
>>>>> you made that design decision so feel free to read it as a pure question.
>>>>>
>>>>
>>>> Understand your point here. The allocation and max_pte_* semantics are
>>>> split between khugepaged-like and fault-like, respectively - which
>>>> could be confusing. Originally, I proposed a MADV_F_COLLAPSE_LIMITS
>>>> flag to control the former's behavior, but agreed to keep things
>>>> simple to start, and expand the interface if/when necessary. I opted
>>>> to ignore max_ptes_* as the default since I envisioned that early
>>>> adopters would "just want it to work". One such example would be
>>>> backing executable text by hugepages on program load when many pages
>>>> haven't been demand-paged in yet.
>>>>
>>>> What do you think?
>>>
>>> I'm just slightly worried that'll make the default MADV_COLLAPSE semantics
>>> blurred.
>>>
>>> To me, a clean default definition for MADV_COLLAPSE would be nice, as "do
>>> khugepaged on this range, and with current thread context".  IMHO any
>>> feature bits then can be supplementing special needs, and I'll take the thp
>>> backing executable example to be one of the (good?) reason we'd need an
>>> extra flag for ignoring the max_ptes_* knobs.
>>>
>>> So personally if I were you maybe I'll start with the simple scheme of that
>>> (even if it won't immediately service a thing) but then add either the
>>> defrag or ignore_max_ptes_* as feature bits later on, with clear use case
>>> descriptions about why we need each of the feature flags.  IMHO numbers
>>> would be even more helpful when there's specific use cases on the show.
>>>
>>> Or, perhaps you think all potential MADV_COLLAPSE users should literally
>>> skip max_ptes_* limitations always?
>>>
>>
>> Thanks for your time and valuable feedback here, Peter. I had a response typed
>> up, but after a few iterations became increasingly unsatisfied with my
>> own response.
>>
>> I think this feature should be able to stand on its own without
>> consideration of a userspace khugepaged, as we have existing concrete
>> examples where it would be useful. In these cases, and I assume almost
>> all other use-cases outside userspace khugepaged, max_ptes_* should be
>> ignored as the fundamental assumption of MADV_COLLAPSE is that the
>> user knows better, and IMHO, khugepaged heuristics shouldn't tell
>> users they are wrong.
> 
> Valid point.  And actually right after I replied I thought similarly on
> whether we need to connect the two interfaces at all..
> 
> It's just that it's very easy to go think like that after reading the cover
> letter since that's exactly what it is comparing to. :)
> 
> There's definitely a difference view on user/kernel level of things, then
> it sounds reasonable to me if we add a new interface it by default has a
> stronger semantics otherwise we may not bother if with MADV_HUGEPAGE's
> existance.
> 
> So maybe max_ptes_* won't even make sense for MADV_COLLAPSE in most cases
> as you said.  And that's a real pure question I asked above, and I feel
> like your answer is actually "yes" we should always ignore the max_ptes_*
> fields until there's a proof that it'll be helpful.
> 
>>
>> But this, as you mention, unsatisfactorily blurs the semantics of
>> MADV_COLLAPSE: "act like khugepaged here, but not here".
>>
>> As such, WDYT about the reverse-side of the coin of what you proposed:
>> to not couple the default behavior of MADV_COLLAPSE with khugepaged at
>> all? I.e. Not tie the allocation semantics to
>> /sys/kernel/mm/transparent_hugepage/khugepaged/defrag. We can add
>> flags as necessary when/if a reimplementation of khugepaged in
>> userspace proves fruitful.
> 
> Let's see whether others have thoughts, but what you proposed here makes
> sense to me.


Makes sense to me. IIUC, the whole handling of max_ptes_* is in place as
 tunable because we don't know what user space is up to.

E.g., have with a very sparse memory layout, we don't want to waste
memory by allocating memory where we actually have no page populated yet
-- could be user space won't reuse that memory in the foreseeable
future. With too many swap entries, we don't want to trigger an
eventually unnecessary overhead of swapping in entries if user space
won't access them in the foreseeable future. Something similar applies
to max_ptes_shared, where one might just end up wasting a lot of memory
eventually in some applications.

So IMHO, with MADV_COLLAPSE we should ignore/disable any heuristics that
try figuring out what user space might be doing. We know exactly what
user space asks for -- and that can be documented properly.

-- 
Thanks,

David / dhildenb

Re: [PATCH v2 00/12] mm: userspace hugepage collapse

[Zach O'Keefe]

On Tue, Apr 19, 2022 at 7:33 AM David Hildenbrand <david@redhat.com> wrote:
>
> On 16.04.22 21:26, Peter Xu wrote:
> > Hi, Zach,
> >
> > On Fri, Apr 15, 2022 at 01:04:04PM -0700, Zach O'Keefe wrote:
> >> On Fri, Apr 15, 2022 at 6:39 AM Peter Xu <peterx@redhat.com> wrote:
> >>>
> >>> On Thu, Apr 14, 2022 at 05:52:43PM -0700, Zach O'Keefe wrote:
> >>>> Hey Peter,
> >>>>
> >>>> Thanks for taking the time to review!
> >>>>
> >>>> On Thu, Apr 14, 2022 at 5:04 PM Peter Xu <peterx@redhat.com> wrote:
> >>>>>
> >>>>> Hi, Zach,
> >>>>>
> >>>>> On Thu, Apr 14, 2022 at 11:06:00AM -0700, Zach O'Keefe wrote:
> >>>>>> process_madvise(2)
> >>>>>>
> >>>>>>       Performs a synchronous collapse of the native pages
> >>>>>>       mapped by the list of iovecs into transparent hugepages.
> >>>>>>
> >>>>>>       Allocation semantics are the same as khugepaged, and depend on
> >>>>>>       (1) the active sysfs settings
> >>>>>>       /sys/kernel/mm/transparent_hugepage/enabled and
> >>>>>>       /sys/kernel/mm/transparent_hugepage/khugepaged/defrag, and (2)
> >>>>>>       the VMA flags of the memory range being collapsed.
> >>>>>>
> >>>>>>       Collapse eligibility criteria differs from khugepaged in that
> >>>>>>       the sysfs files
> >>>>>>       /sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_[none|swap|shared]
> >>>>>>       are ignored.
> >>>>>
> >>>>> The userspace khugepaged idea definitely makes sense to me, though I'm
> >>>>> curious how the line is drown on the different behaviors here by explicitly
> >>>>> ignoring the max_ptes_* entries.
> >>>>>
> >>>>> Let's assume the initiative is to duplicate a more data-aware khugepaged in
> >>>>> the userspace, then IMHO it makes more sense to start with all the policies
> >>>>> that applies to khugepaged already, including max_pte_*.
> >>>>>
> >>>>> I can understand the willingness to provide even stronger semantics here
> >>>>> than khugepaged since the userspace could have very clear knowledge of how
> >>>>> to provision the memories (better than a kernel scanner).  It's just that
> >>>>> IMHO it could be slightly confusing if the new interface only partially
> >>>>> apply the khugepaged rules.
> >>>>>
> >>>>> No strong opinion here.  It could already been a trade-off after the
> >>>>> discussion from the RFC with Michal which I read..  Just curious about how
> >>>>> you made that design decision so feel free to read it as a pure question.
> >>>>>
> >>>>
> >>>> Understand your point here. The allocation and max_pte_* semantics are
> >>>> split between khugepaged-like and fault-like, respectively - which
> >>>> could be confusing. Originally, I proposed a MADV_F_COLLAPSE_LIMITS
> >>>> flag to control the former's behavior, but agreed to keep things
> >>>> simple to start, and expand the interface if/when necessary. I opted
> >>>> to ignore max_ptes_* as the default since I envisioned that early
> >>>> adopters would "just want it to work". One such example would be
> >>>> backing executable text by hugepages on program load when many pages
> >>>> haven't been demand-paged in yet.
> >>>>
> >>>> What do you think?
> >>>
> >>> I'm just slightly worried that'll make the default MADV_COLLAPSE semantics
> >>> blurred.
> >>>
> >>> To me, a clean default definition for MADV_COLLAPSE would be nice, as "do
> >>> khugepaged on this range, and with current thread context".  IMHO any
> >>> feature bits then can be supplementing special needs, and I'll take the thp
> >>> backing executable example to be one of the (good?) reason we'd need an
> >>> extra flag for ignoring the max_ptes_* knobs.
> >>>
> >>> So personally if I were you maybe I'll start with the simple scheme of that
> >>> (even if it won't immediately service a thing) but then add either the
> >>> defrag or ignore_max_ptes_* as feature bits later on, with clear use case
> >>> descriptions about why we need each of the feature flags.  IMHO numbers
> >>> would be even more helpful when there's specific use cases on the show.
> >>>
> >>> Or, perhaps you think all potential MADV_COLLAPSE users should literally
> >>> skip max_ptes_* limitations always?
> >>>
> >>
> >> Thanks for your time and valuable feedback here, Peter. I had a response typed
> >> up, but after a few iterations became increasingly unsatisfied with my
> >> own response.
> >>
> >> I think this feature should be able to stand on its own without
> >> consideration of a userspace khugepaged, as we have existing concrete
> >> examples where it would be useful. In these cases, and I assume almost
> >> all other use-cases outside userspace khugepaged, max_ptes_* should be
> >> ignored as the fundamental assumption of MADV_COLLAPSE is that the
> >> user knows better, and IMHO, khugepaged heuristics shouldn't tell
> >> users they are wrong.
> >
> > Valid point.  And actually right after I replied I thought similarly on
> > whether we need to connect the two interfaces at all..
> >
> > It's just that it's very easy to go think like that after reading the cover
> > letter since that's exactly what it is comparing to. :)
> >
> > There's definitely a difference view on user/kernel level of things, then
> > it sounds reasonable to me if we add a new interface it by default has a
> > stronger semantics otherwise we may not bother if with MADV_HUGEPAGE's
> > existance.
> >
> > So maybe max_ptes_* won't even make sense for MADV_COLLAPSE in most cases
> > as you said.  And that's a real pure question I asked above, and I feel
> > like your answer is actually "yes" we should always ignore the max_ptes_*
> > fields until there's a proof that it'll be helpful.
> >
> >>
> >> But this, as you mention, unsatisfactorily blurs the semantics of
> >> MADV_COLLAPSE: "act like khugepaged here, but not here".
> >>
> >> As such, WDYT about the reverse-side of the coin of what you proposed:
> >> to not couple the default behavior of MADV_COLLAPSE with khugepaged at
> >> all? I.e. Not tie the allocation semantics to
> >> /sys/kernel/mm/transparent_hugepage/khugepaged/defrag. We can add
> >> flags as necessary when/if a reimplementation of khugepaged in
> >> userspace proves fruitful.
> >
> > Let's see whether others have thoughts, but what you proposed here makes
> > sense to me.
>
>
Hey David - thanks again for taking the time to read and comment.

> Makes sense to me. IIUC, the whole handling of max_ptes_* is in place as
>  tunable because we don't know what user space is up to.
>

Agreed.

> E.g., have with a very sparse memory layout, we don't want to waste
> memory by allocating memory where we actually have no page populated yet
> -- could be user space won't reuse that memory in the foreseeable
> future. With too many swap entries, we don't want to trigger an
> eventually unnecessary overhead of swapping in entries if user space
> won't access them in the foreseeable future. Something similar applies
> to max_ptes_shared, where one might just end up wasting a lot of memory
> eventually in some applications.
>
> So IMHO, with MADV_COLLAPSE we should ignore/disable any heuristics that
> try figuring out what user space might be doing. We know exactly what
> user space asks for -- and that can be documented properly.
>

Sounds good to me. Would you also be in favor of decoupling allocation
semantics from khugepaged? I.e. we'll pick some default gfp flags and
not depend on /sys/kernel/mm/transparent_hugepage/khugepaged/defrag?

Thanks again,

Zach

> --
> Thanks,
>
> David / dhildenb
>

Re: [PATCH v2 00/12] mm: userspace hugepage collapse

[David Hildenbrand]

>> E.g., have with a very sparse memory layout, we don't want to waste
>> memory by allocating memory where we actually have no page populated yet
>> -- could be user space won't reuse that memory in the foreseeable
>> future. With too many swap entries, we don't want to trigger an
>> eventually unnecessary overhead of swapping in entries if user space
>> won't access them in the foreseeable future. Something similar applies
>> to max_ptes_shared, where one might just end up wasting a lot of memory
>> eventually in some applications.
>>
>> So IMHO, with MADV_COLLAPSE we should ignore/disable any heuristics that
>> try figuring out what user space might be doing. We know exactly what
>> user space asks for -- and that can be documented properly.
>>

Just a thought, if we ever want to implement khugepaged in user space,
it could theoretically obtain similar information using e.g., the
pagemap. It wouldn't be race-free, but the question is if it would matter.

I consider the primary use case of giving an application more precise
control over actual THP placement.

> 
> Sounds good to me. Would you also be in favor of decoupling allocation
> semantics from khugepaged? I.e. we'll pick some default gfp flags and
> not depend on /sys/kernel/mm/transparent_hugepage/khugepaged/defrag?

Good question. It's not really a heuristic like that other stuff.

Easy answer: we're not dealing with khugepaged, so anything in
/sys/kernel/mm/transparent_hugepage/khugepaged/ shouldn't apply?

Sure, we could have a separate toggles for MADV_COLLAPSE.

Maybe we simply want a dedicated syscall where we can specify additional
options ... but maybe that simply over-complicates the problem.

-- 
Thanks,

David / dhildenb

Re: [PATCH v2 00/12] mm: userspace hugepage collapse

[Zach O'Keefe]

On Tue, Apr 19, 2022 at 1:03 PM David Hildenbrand <david@redhat.com> wrote:
>
> >> E.g., have with a very sparse memory layout, we don't want to waste
> >> memory by allocating memory where we actually have no page populated yet
> >> -- could be user space won't reuse that memory in the foreseeable
> >> future. With too many swap entries, we don't want to trigger an
> >> eventually unnecessary overhead of swapping in entries if user space
> >> won't access them in the foreseeable future. Something similar applies
> >> to max_ptes_shared, where one might just end up wasting a lot of memory
> >> eventually in some applications.
> >>
> >> So IMHO, with MADV_COLLAPSE we should ignore/disable any heuristics that
> >> try figuring out what user space might be doing. We know exactly what
> >> user space asks for -- and that can be documented properly.
> >>
>
> Just a thought, if we ever want to implement khugepaged in user space,
> it could theoretically obtain similar information using e.g., the
> pagemap. It wouldn't be race-free, but the question is if it would matter.
>
> I consider the primary use case of giving an application more precise
> control over actual THP placement.
>

Good point about the pagemap and agree about the primary use case -
I'll make that clear in v3 cover letter.

> >
> > Sounds good to me. Would you also be in favor of decoupling allocation
> > semantics from khugepaged? I.e. we'll pick some default gfp flags and
> > not depend on /sys/kernel/mm/transparent_hugepage/khugepaged/defrag?
>
> Good question. It's not really a heuristic like that other stuff.
>
> Easy answer: we're not dealing with khugepaged, so anything in
> /sys/kernel/mm/transparent_hugepage/khugepaged/ shouldn't apply?
>

That's what I'm thinking now too. If there's no objections, I'll
proceed in that direction for v3.

> Sure, we could have a separate toggles for MADV_COLLAPSE.
>
> Maybe we simply want a dedicated syscall where we can specify additional
> options ... but maybe that simply over-complicates the problem.
>

Thankfully process_madvise(2) has flags, and madvise(2) users can
always migrate to using process_madvise(2) on self. Piggy-backing off
madvise infrastructure for these "non-advice actions" (e.g.
MADV_PAGEOUT) seems to be the norm.

Thanks as always for your time and thoughts!

Zach

> --
> Thanks,
>
> David / dhildenb
>

Re: [PATCH v2 00/12] mm: userspace hugepage collapse

[Yang Shi]

On Tue, Apr 19, 2022 at 3:43 PM Zach O'Keefe <zokeefe@google.com> wrote:
>
> On Tue, Apr 19, 2022 at 1:03 PM David Hildenbrand <david@redhat.com> wrote:
> >
> > >> E.g., have with a very sparse memory layout, we don't want to waste
> > >> memory by allocating memory where we actually have no page populated yet
> > >> -- could be user space won't reuse that memory in the foreseeable
> > >> future. With too many swap entries, we don't want to trigger an
> > >> eventually unnecessary overhead of swapping in entries if user space
> > >> won't access them in the foreseeable future. Something similar applies
> > >> to max_ptes_shared, where one might just end up wasting a lot of memory
> > >> eventually in some applications.
> > >>
> > >> So IMHO, with MADV_COLLAPSE we should ignore/disable any heuristics that
> > >> try figuring out what user space might be doing. We know exactly what
> > >> user space asks for -- and that can be documented properly.
> > >>
> >
> > Just a thought, if we ever want to implement khugepaged in user space,
> > it could theoretically obtain similar information using e.g., the
> > pagemap. It wouldn't be race-free, but the question is if it would matter.
> >
> > I consider the primary use case of giving an application more precise
> > control over actual THP placement.
> >
>
> Good point about the pagemap and agree about the primary use case -
> I'll make that clear in v3 cover letter.
>
> > >
> > > Sounds good to me. Would you also be in favor of decoupling allocation
> > > semantics from khugepaged? I.e. we'll pick some default gfp flags and
> > > not depend on /sys/kernel/mm/transparent_hugepage/khugepaged/defrag?
> >
> > Good question. It's not really a heuristic like that other stuff.
> >
> > Easy answer: we're not dealing with khugepaged, so anything in
> > /sys/kernel/mm/transparent_hugepage/khugepaged/ shouldn't apply?
> >
>
> That's what I'm thinking now too. If there's no objections, I'll
> proceed in that direction for v3.

I agree, we should not treat MADV_COLLAPSE as "userspace khugepaged"
IMHO. It is still best effort though, but it is requested by the users
explicitly so kernel should trust the users' judgement and ignore
those max_ptes_* since we should assume the users know what they are
doing and the cost.

>
> > Sure, we could have a separate toggles for MADV_COLLAPSE.
> >
> > Maybe we simply want a dedicated syscall where we can specify additional
> > options ... but maybe that simply over-complicates the problem.
> >
>
> Thankfully process_madvise(2) has flags, and madvise(2) users can
> always migrate to using process_madvise(2) on self. Piggy-backing off
> madvise infrastructure for these "non-advice actions" (e.g.
> MADV_PAGEOUT) seems to be the norm.
>
> Thanks as always for your time and thoughts!
>
> Zach
>
> > --
> > Thanks,
> >
> > David / dhildenb
> >

Re: [PATCH v2 00/12] mm: userspace hugepage collapse

[Zach O'Keefe]

On Wed, Apr 20, 2022 at 5:57 PM Yang Shi <shy828301@gmail.com> wrote:
>
> On Tue, Apr 19, 2022 at 3:43 PM Zach O'Keefe <zokeefe@google.com> wrote:
> >
> > On Tue, Apr 19, 2022 at 1:03 PM David Hildenbrand <david@redhat.com> wrote:
> > >
> > > >> E.g., have with a very sparse memory layout, we don't want to waste
> > > >> memory by allocating memory where we actually have no page populated yet
> > > >> -- could be user space won't reuse that memory in the foreseeable
> > > >> future. With too many swap entries, we don't want to trigger an
> > > >> eventually unnecessary overhead of swapping in entries if user space
> > > >> won't access them in the foreseeable future. Something similar applies
> > > >> to max_ptes_shared, where one might just end up wasting a lot of memory
> > > >> eventually in some applications.
> > > >>
> > > >> So IMHO, with MADV_COLLAPSE we should ignore/disable any heuristics that
> > > >> try figuring out what user space might be doing. We know exactly what
> > > >> user space asks for -- and that can be documented properly.
> > > >>
> > >
> > > Just a thought, if we ever want to implement khugepaged in user space,
> > > it could theoretically obtain similar information using e.g., the
> > > pagemap. It wouldn't be race-free, but the question is if it would matter.
> > >
> > > I consider the primary use case of giving an application more precise
> > > control over actual THP placement.
> > >
> >
> > Good point about the pagemap and agree about the primary use case -
> > I'll make that clear in v3 cover letter.
> >
> > > >
> > > > Sounds good to me. Would you also be in favor of decoupling allocation
> > > > semantics from khugepaged? I.e. we'll pick some default gfp flags and
> > > > not depend on /sys/kernel/mm/transparent_hugepage/khugepaged/defrag?
> > >
> > > Good question. It's not really a heuristic like that other stuff.
> > >
> > > Easy answer: we're not dealing with khugepaged, so anything in
> > > /sys/kernel/mm/transparent_hugepage/khugepaged/ shouldn't apply?
> > >
> >
> > That's what I'm thinking now too. If there's no objections, I'll
> > proceed in that direction for v3.
>
> I agree, we should not treat MADV_COLLAPSE as "userspace khugepaged"
> IMHO. It is still best effort though, but it is requested by the users
> explicitly so kernel should trust the users' judgement and ignore
> those max_ptes_* since we should assume the users know what they are
> doing and the cost.
>

Thanks for reading and giving your thoughts, Yang. Glad to hear we are
aligned here!

I'll send out a v3 early next week. Only real change is the gfp flags,
but I want to avoid spamming folks so soon since v2.

Thanks,
Zach

> >
> > > Sure, we could have a separate toggles for MADV_COLLAPSE.
> > >
> > > Maybe we simply want a dedicated syscall where we can specify additional
> > > options ... but maybe that simply over-complicates the problem.
> > >
> >
> > Thankfully process_madvise(2) has flags, and madvise(2) users can
> > always migrate to using process_madvise(2) on self. Piggy-backing off
> > madvise infrastructure for these "non-advice actions" (e.g.
> > MADV_PAGEOUT) seems to be the norm.
> >
> > Thanks as always for your time and thoughts!
> >
> > Zach
> >
> > > --
> > > Thanks,
> > >
> > > David / dhildenb
> > >