[PATCH v4 0/2] khugepaged: collapse pmd for pte-mapped THP

[PATCH v4 0/2] khugepaged: collapse pmd for pte-mapped THP

[Song Liu]

Changes v3 => v4:
1. Simplify locking for pte_mapped_thp (Oleg).
2. Improve checks for the page in collapse_pte_mapped_thp() (Oleg).
3. Move HPAGE_PMD_MASK to collapse_pte_mapped_thp() (kbuild test robot).

Changes v2 => v3:
1. Update vma/pmd check in collapse_pte_mapped_thp() (Oleg).
2. Add Acked-by from Kirill

Changes v1 => v2:
1. Call collapse_pte_mapped_thp() directly from uprobe_write_opcode();
2. Add VM_BUG_ON() for addr alignment in khugepaged_add_pte_mapped_thp()
   and collapse_pte_mapped_thp().

This set is the newer version of 5/6 and 6/6 of [1]. Newer version of
1-4 of the work [2] was recently picked by Andrew.

Patch 1 enables khugepaged to handle pte-mapped THP. These THPs are left
in such state when khugepaged failed to get exclusive lock of mmap_sem.

Patch 2 leverages work in 1 for uprobe on THP. After [2], uprobe only
splits the PMD. When the uprobe is disabled, we get pte-mapped THP.
After this set, these pte-mapped THP will be collapsed as pmd-mapped.

[1] https://lkml.org/lkml/2019/6/23/23
[2] https://www.spinics.net/lists/linux-mm/msg185889.html

Song Liu (2):
  khugepaged: enable collapse pmd for pte-mapped THP
  uprobe: collapse THP pmd after removing all uprobes

 include/linux/khugepaged.h |  12 ++++
 kernel/events/uprobes.c    |   9 +++
 mm/khugepaged.c            | 125 ++++++++++++++++++++++++++++++++++++-
 3 files changed, 145 insertions(+), 1 deletion(-)

--
2.17.1

[PATCH v4 1/2] khugepaged: enable collapse pmd for pte-mapped THP

[Song Liu]

khugepaged needs exclusive mmap_sem to access page table. When it fails
to lock mmap_sem, the page will fault in as pte-mapped THP. As the page
is already a THP, khugepaged will not handle this pmd again.

This patch enables the khugepaged to retry collapse the page table.

struct mm_slot (in khugepaged.c) is extended with an array, containing
addresses of pte-mapped THPs. We use array here for simplicity. We can
easily replace it with more advanced data structures when needed.

In khugepaged_scan_mm_slot(), if the mm contains pte-mapped THP, we try
to collapse the page table.

Since collapse may happen at an later time, some pages may already fault
in. collapse_pte_mapped_thp() is added to properly handle these pages.
collapse_pte_mapped_thp() also double checks whether all ptes in this pmd
are mapping to the same THP. This is necessary because some subpage of
the THP may be replaced, for example by uprobe. In such cases, it is not
possible to collapse the pmd.

Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Song Liu <songliubraving@fb.com>
---
 include/linux/khugepaged.h |  12 ++++
 mm/khugepaged.c            | 123 ++++++++++++++++++++++++++++++++++++-
 2 files changed, 134 insertions(+), 1 deletion(-)

diff --git a/include/linux/khugepaged.h b/include/linux/khugepaged.h
index 082d1d2a5216..bc45ea1efbf7 100644
--- a/include/linux/khugepaged.h
+++ b/include/linux/khugepaged.h
@@ -15,6 +15,14 @@ extern int __khugepaged_enter(struct mm_struct *mm);
 extern void __khugepaged_exit(struct mm_struct *mm);
 extern int khugepaged_enter_vma_merge(struct vm_area_struct *vma,
 				      unsigned long vm_flags);
+#ifdef CONFIG_SHMEM
+extern void collapse_pte_mapped_thp(struct mm_struct *mm, unsigned long addr);
+#else
+static inline void collapse_pte_mapped_thp(struct mm_struct *mm,
+					   unsigned long addr)
+{
+}
+#endif
 
 #define khugepaged_enabled()					       \
 	(transparent_hugepage_flags &				       \
@@ -73,6 +81,10 @@ static inline int khugepaged_enter_vma_merge(struct vm_area_struct *vma,
 {
 	return 0;
 }
+static inline void collapse_pte_mapped_thp(struct mm_struct *mm,
+					   unsigned long addr)
+{
+}
 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
 
 #endif /* _LINUX_KHUGEPAGED_H */
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index eaaa21b23215..ba36ff5c1d82 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -76,6 +76,8 @@ static __read_mostly DEFINE_HASHTABLE(mm_slots_hash, MM_SLOTS_HASH_BITS);
 
 static struct kmem_cache *mm_slot_cache __read_mostly;
 
+#define MAX_PTE_MAPPED_THP 8
+
 /**
  * struct mm_slot - hash lookup from mm to mm_slot
  * @hash: hash collision list
@@ -86,6 +88,10 @@ struct mm_slot {
 	struct hlist_node hash;
 	struct list_head mm_node;
 	struct mm_struct *mm;
+
+	/* pte-mapped THP in this mm */
+	int nr_pte_mapped_thp;
+	unsigned long pte_mapped_thp[MAX_PTE_MAPPED_THP];
 };
 
 /**
@@ -1248,6 +1254,119 @@ static void collect_mm_slot(struct mm_slot *mm_slot)
 }
 
 #if defined(CONFIG_SHMEM) && defined(CONFIG_TRANSPARENT_HUGE_PAGECACHE)
+/*
+ * Notify khugepaged that given addr of the mm is pte-mapped THP. Then
+ * khugepaged should try to collapse the page table.
+ */
+static int khugepaged_add_pte_mapped_thp(struct mm_struct *mm,
+					 unsigned long addr)
+{
+	struct mm_slot *mm_slot;
+
+	VM_BUG_ON(addr & ~HPAGE_PMD_MASK);
+
+	spin_lock(&khugepaged_mm_lock);
+	mm_slot = get_mm_slot(mm);
+	if (likely(mm_slot && mm_slot->nr_pte_mapped_thp < MAX_PTE_MAPPED_THP))
+		mm_slot->pte_mapped_thp[mm_slot->nr_pte_mapped_thp++] = addr;
+	spin_unlock(&khugepaged_mm_lock);
+	return 0;
+}
+
+/**
+ * Try to collapse a pte-mapped THP for mm at address haddr.
+ *
+ * This function checks whether all the PTEs in the PMD are pointing to the
+ * right THP. If so, retract the page table so the THP can refault in with
+ * as pmd-mapped.
+ */
+void collapse_pte_mapped_thp(struct mm_struct *mm, unsigned long addr)
+{
+	unsigned long haddr = addr & HPAGE_PMD_MASK;
+	struct vm_area_struct *vma = find_vma(mm, haddr);
+	pmd_t *pmd = mm_find_pmd(mm, haddr);
+	struct page *hpage = NULL;
+	spinlock_t *ptl;
+	int count = 0;
+	pmd_t _pmd;
+	int i;
+
+	if (!vma || !vma->vm_file || !pmd ||
+	    vma->vm_start > haddr || vma->vm_end < haddr + HPAGE_PMD_SIZE)
+		return;
+
+	/* step 1: check all mapped PTEs are to the right huge page */
+	for (i = 0, addr = haddr; i < HPAGE_PMD_NR; i++, addr += PAGE_SIZE) {
+		pte_t *pte = pte_offset_map(pmd, addr);
+		struct page *page;
+
+		if (pte_none(*pte))
+			continue;
+
+		page = vm_normal_page(vma, addr, *pte);
+
+		if (!page || !PageCompound(page))
+			return;
+
+		if (!hpage) {
+			hpage = compound_head(page);
+			if (hpage->mapping != vma->vm_file->f_mapping)
+				return;
+		}
+
+		if (hpage + i != page)
+			return;
+		count++;
+	}
+
+	/* step 2: adjust rmap */
+	for (i = 0, addr = haddr; i < HPAGE_PMD_NR; i++, addr += PAGE_SIZE) {
+		pte_t *pte = pte_offset_map(pmd, addr);
+		struct page *page;
+
+		if (pte_none(*pte))
+			continue;
+		page = vm_normal_page(vma, addr, *pte);
+		page_remove_rmap(page, false);
+	}
+
+	/* step 3: set proper refcount and mm_counters. */
+	if (hpage) {
+		page_ref_sub(hpage, count);
+		add_mm_counter(vma->vm_mm, mm_counter_file(hpage), -count);
+	}
+
+	/* step 4: collapse pmd */
+	ptl = pmd_lock(vma->vm_mm, pmd);
+	_pmd = pmdp_collapse_flush(vma, addr, pmd);
+	spin_unlock(ptl);
+	mm_dec_nr_ptes(mm);
+	pte_free(mm, pmd_pgtable(_pmd));
+}
+
+static int khugepaged_collapse_pte_mapped_thps(struct mm_slot *mm_slot)
+{
+	struct mm_struct *mm = mm_slot->mm;
+	int i;
+
+	if (likely(mm_slot->nr_pte_mapped_thp == 0))
+		return 0;
+
+	if (!down_write_trylock(&mm->mmap_sem))
+		return -EBUSY;
+
+	if (unlikely(khugepaged_test_exit(mm)))
+		goto out;
+
+	for (i = 0; i < mm_slot->nr_pte_mapped_thp; i++)
+		collapse_pte_mapped_thp(mm, mm_slot->pte_mapped_thp[i]);
+
+out:
+	mm_slot->nr_pte_mapped_thp = 0;
+	up_write(&mm->mmap_sem);
+	return 0;
+}
+
 static void retract_page_tables(struct address_space *mapping, pgoff_t pgoff)
 {
 	struct vm_area_struct *vma;
@@ -1281,7 +1400,8 @@ static void retract_page_tables(struct address_space *mapping, pgoff_t pgoff)
 			up_write(&vma->vm_mm->mmap_sem);
 			mm_dec_nr_ptes(vma->vm_mm);
 			pte_free(vma->vm_mm, pmd_pgtable(_pmd));
-		}
+		} else
+			khugepaged_add_pte_mapped_thp(vma->vm_mm, addr);
 	}
 	i_mmap_unlock_write(mapping);
 }
@@ -1668,6 +1788,7 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
 		khugepaged_scan.mm_slot = mm_slot;
 	}
 	spin_unlock(&khugepaged_mm_lock);
+	khugepaged_collapse_pte_mapped_thps(mm_slot);
 
 	mm = mm_slot->mm;
 	/*
-- 
2.17.1

Re: [PATCH v4 1/2] khugepaged: enable collapse pmd for pte-mapped THP

[Oleg Nesterov]

On 08/02, Song Liu wrote:
>
> +void collapse_pte_mapped_thp(struct mm_struct *mm, unsigned long addr)
> +{
> +	unsigned long haddr = addr & HPAGE_PMD_MASK;
> +	struct vm_area_struct *vma = find_vma(mm, haddr);
> +	pmd_t *pmd = mm_find_pmd(mm, haddr);
> +	struct page *hpage = NULL;
> +	spinlock_t *ptl;
> +	int count = 0;
> +	pmd_t _pmd;
> +	int i;
> +
> +	if (!vma || !vma->vm_file || !pmd ||
> +	    vma->vm_start > haddr || vma->vm_end < haddr + HPAGE_PMD_SIZE)
> +		return;

I still can't understand why is it safe to blindly use mm_find_pmd().

Say, what pmd_offset(pud, address) will return to this function if
pud_huge() == T? IIUC, this is possible if is_file_hugepages(vm_file).
How the code below can use this result?

I think you need something like hugepage_vma_check() or even
hugepage_vma_revalidate().

Oleg.

Re: [PATCH v4 1/2] khugepaged: enable collapse pmd for pte-mapped THP

[Song Liu]

> On Aug 6, 2019, at 3:02 AM, Oleg Nesterov <oleg@redhat.com> wrote:
> 
> On 08/02, Song Liu wrote:
>> 
>> +void collapse_pte_mapped_thp(struct mm_struct *mm, unsigned long addr)
>> +{
>> +	unsigned long haddr = addr & HPAGE_PMD_MASK;
>> +	struct vm_area_struct *vma = find_vma(mm, haddr);
>> +	pmd_t *pmd = mm_find_pmd(mm, haddr);
>> +	struct page *hpage = NULL;
>> +	spinlock_t *ptl;
>> +	int count = 0;
>> +	pmd_t _pmd;
>> +	int i;
>> +
>> +	if (!vma || !vma->vm_file || !pmd ||
>> +	    vma->vm_start > haddr || vma->vm_end < haddr + HPAGE_PMD_SIZE)
>> +		return;
> 
> I still can't understand why is it safe to blindly use mm_find_pmd().
> 
> Say, what pmd_offset(pud, address) will return to this function if
> pud_huge() == T? IIUC, this is possible if is_file_hugepages(vm_file).
> How the code below can use this result?

IIUC, the concern is matching files in hugetlbfs. Maybe we can exclude
that specifically? 

> 
> I think you need something like hugepage_vma_check() or even
> hugepage_vma_revalidate().

To use hugepage_vma_check(), we will need to set VM_HUGEPAGE for the 
following case:

    mount shm with huge=always
    copy app to shm
    start app and enable uprobe
    disable uprobe

This vma will not have VM_HUGEPAGE, so it will fail hugepage_vma_check().

How about something like:

diff --git i/kernel/events/uprobes.c w/kernel/events/uprobes.c
index 94d38a39d72e..f0d3e367f907 100644
--- i/kernel/events/uprobes.c
+++ w/kernel/events/uprobes.c
@@ -532,8 +532,10 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
                                put_page(new_page);
                                new_page = NULL;

-                               if (PageCompound(orig_page))
+                               if (PageCompound(orig_page)) {
                                        orig_page_huge = true;
+                                       vma->vm_flags |= VM_HUGEPAGE;
+                               }
                        }
                        put_page(orig_page);
                }

Thanks,
Song

[PATCH v4 2/2] uprobe: collapse THP pmd after removing all uprobes

[Song Liu]

After all uprobes are removed from the huge page (with PTE pgtable), it
is possible to collapse the pmd and benefit from THP again. This patch
does the collapse by calling collapse_pte_mapped_thp().

Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Song Liu <songliubraving@fb.com>
---
 kernel/events/uprobes.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
index 27b596f14463..94d38a39d72e 100644
--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -26,6 +26,7 @@
 #include <linux/percpu-rwsem.h>
 #include <linux/task_work.h>
 #include <linux/shmem_fs.h>
+#include <linux/khugepaged.h>
 
 #include <linux/uprobes.h>
 
@@ -472,6 +473,7 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
 	struct page *old_page, *new_page;
 	struct vm_area_struct *vma;
 	int ret, is_register, ref_ctr_updated = 0;
+	bool orig_page_huge = false;
 
 	is_register = is_swbp_insn(&opcode);
 	uprobe = container_of(auprobe, struct uprobe, arch);
@@ -529,6 +531,9 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
 				/* let go new_page */
 				put_page(new_page);
 				new_page = NULL;
+
+				if (PageCompound(orig_page))
+					orig_page_huge = true;
 			}
 			put_page(orig_page);
 		}
@@ -547,6 +552,10 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
 	if (ret && is_register && ref_ctr_updated)
 		update_ref_ctr(uprobe, mm, -1);
 
+	/* try collapse pmd for compound page */
+	if (!ret && orig_page_huge)
+		collapse_pte_mapped_thp(mm, vaddr);
+
 	return ret;
 }
 
-- 
2.17.1