Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
[linux-3.10.git] / mm / ksm.c
index d9e3cfc..5157385 100644 (file)
--- a/mm/ksm.c
+++ b/mm/ksm.c
 #include <linux/wait.h>
 #include <linux/slab.h>
 #include <linux/rbtree.h>
+#include <linux/memory.h>
 #include <linux/mmu_notifier.h>
+#include <linux/swap.h>
 #include <linux/ksm.h>
+#include <linux/hash.h>
+#include <linux/freezer.h>
+#include <linux/oom.h>
 
 #include <asm/tlbflush.h>
+#include "internal.h"
 
 /*
  * A few notes about the KSM scanning process,
  * struct mm_slot - ksm information per mm that is being scanned
  * @link: link to the mm_slots hash list
  * @mm_list: link into the mm_slots list, rooted in ksm_mm_head
- * @rmap_list: head for this mm_slot's list of rmap_items
+ * @rmap_list: head for this mm_slot's singly-linked list of rmap_items
  * @mm: the mm that this information is valid for
  */
 struct mm_slot {
        struct hlist_node link;
        struct list_head mm_list;
-       struct list_head rmap_list;
+       struct rmap_item *rmap_list;
        struct mm_struct *mm;
 };
 
@@ -92,7 +98,7 @@ struct mm_slot {
  * struct ksm_scan - cursor for scanning
  * @mm_slot: the current mm_slot we are scanning
  * @address: the next address inside that to be scanned
- * @rmap_item: the current rmap that we are scanning inside the rmap_list
+ * @rmap_list: link to the next rmap to be scanned in the rmap_list
  * @seqnr: count of completed full scans (needed when removing unstable node)
  *
  * There is only the one ksm_scan instance of this cursor structure.
@@ -100,44 +106,59 @@ struct mm_slot {
 struct ksm_scan {
        struct mm_slot *mm_slot;
        unsigned long address;
-       struct rmap_item *rmap_item;
+       struct rmap_item **rmap_list;
        unsigned long seqnr;
 };
 
 /**
+ * struct stable_node - node of the stable rbtree
+ * @node: rb node of this ksm page in the stable tree
+ * @hlist: hlist head of rmap_items using this ksm page
+ * @kpfn: page frame number of this ksm page
+ */
+struct stable_node {
+       struct rb_node node;
+       struct hlist_head hlist;
+       unsigned long kpfn;
+};
+
+/**
  * struct rmap_item - reverse mapping item for virtual addresses
- * @link: link into mm_slot's rmap_list (rmap_list is per mm)
+ * @rmap_list: next rmap_item in mm_slot's singly-linked rmap_list
+ * @anon_vma: pointer to anon_vma for this mm,address, when in stable tree
  * @mm: the memory structure this rmap_item is pointing into
  * @address: the virtual address this rmap_item tracks (+ flags in low bits)
  * @oldchecksum: previous checksum of the page at that virtual address
- * @node: rb_node of this rmap_item in either unstable or stable tree
- * @next: next rmap_item hanging off the same node of the stable tree
- * @prev: previous rmap_item hanging off the same node of the stable tree
+ * @node: rb node of this rmap_item in the unstable tree
+ * @head: pointer to stable_node heading this list in the stable tree
+ * @hlist: link into hlist of rmap_items hanging off that stable_node
  */
 struct rmap_item {
-       struct list_head link;
+       struct rmap_item *rmap_list;
+       struct anon_vma *anon_vma;      /* when stable */
        struct mm_struct *mm;
        unsigned long address;          /* + low bits used for flags below */
+       unsigned int oldchecksum;       /* when unstable */
        union {
-               unsigned int oldchecksum;               /* when unstable */
-               struct rmap_item *next;                 /* when stable */
-       };
-       union {
-               struct rb_node node;                    /* when tree node */
-               struct rmap_item *prev;                 /* in stable list */
+               struct rb_node node;    /* when node of unstable tree */
+               struct {                /* when listed from stable tree */
+                       struct stable_node *head;
+                       struct hlist_node hlist;
+               };
        };
 };
 
 #define SEQNR_MASK     0x0ff   /* low bits of unstable tree seqnr */
-#define NODE_FLAG      0x100   /* is a node of unstable or stable tree */
-#define STABLE_FLAG    0x200   /* is a node or list item of stable tree */
+#define UNSTABLE_FLAG  0x100   /* is a node of the unstable tree */
+#define STABLE_FLAG    0x200   /* is listed from the stable tree */
 
 /* The stable and unstable tree heads */
 static struct rb_root root_stable_tree = RB_ROOT;
 static struct rb_root root_unstable_tree = RB_ROOT;
 
-#define MM_SLOTS_HASH_HEADS 1024
-static struct hlist_head *mm_slots_hash;
+#define MM_SLOTS_HASH_SHIFT 10
+#define MM_SLOTS_HASH_HEADS (1 << MM_SLOTS_HASH_SHIFT)
+static struct hlist_head mm_slots_hash[MM_SLOTS_HASH_HEADS];
 
 static struct mm_slot ksm_mm_head = {
        .mm_list = LIST_HEAD_INIT(ksm_mm_head.mm_list),
@@ -147,6 +168,7 @@ static struct ksm_scan ksm_scan = {
 };
 
 static struct kmem_cache *rmap_item_cache;
+static struct kmem_cache *stable_node_cache;
 static struct kmem_cache *mm_slot_cache;
 
 /* The number of nodes in the stable tree */
@@ -161,19 +183,16 @@ static unsigned long ksm_pages_unshared;
 /* The number of rmap_items in use: to calculate pages_volatile */
 static unsigned long ksm_rmap_items;
 
-/* Limit on the number of unswappable pages used */
-static unsigned long ksm_max_kernel_pages;
-
 /* Number of pages ksmd should scan in one batch */
-static unsigned int ksm_thread_pages_to_scan;
+static unsigned int ksm_thread_pages_to_scan = 100;
 
 /* Milliseconds ksmd should sleep between batches */
-static unsigned int ksm_thread_sleep_millisecs;
+static unsigned int ksm_thread_sleep_millisecs = 20;
 
 #define KSM_RUN_STOP   0
 #define KSM_RUN_MERGE  1
 #define KSM_RUN_UNMERGE        2
-static unsigned int ksm_run;
+static unsigned int ksm_run = KSM_RUN_STOP;
 
 static DECLARE_WAIT_QUEUE_HEAD(ksm_thread_wait);
 static DEFINE_MUTEX(ksm_thread_mutex);
@@ -189,13 +208,19 @@ static int __init ksm_slab_init(void)
        if (!rmap_item_cache)
                goto out;
 
+       stable_node_cache = KSM_KMEM_CACHE(stable_node, 0);
+       if (!stable_node_cache)
+               goto out_free1;
+
        mm_slot_cache = KSM_KMEM_CACHE(mm_slot, 0);
        if (!mm_slot_cache)
-               goto out_free;
+               goto out_free2;
 
        return 0;
 
-out_free:
+out_free2:
+       kmem_cache_destroy(stable_node_cache);
+out_free1:
        kmem_cache_destroy(rmap_item_cache);
 out:
        return -ENOMEM;
@@ -204,6 +229,7 @@ out:
 static void __init ksm_slab_free(void)
 {
        kmem_cache_destroy(mm_slot_cache);
+       kmem_cache_destroy(stable_node_cache);
        kmem_cache_destroy(rmap_item_cache);
        mm_slot_cache = NULL;
 }
@@ -225,30 +251,26 @@ static inline void free_rmap_item(struct rmap_item *rmap_item)
        kmem_cache_free(rmap_item_cache, rmap_item);
 }
 
-static inline struct mm_slot *alloc_mm_slot(void)
+static inline struct stable_node *alloc_stable_node(void)
 {
-       if (!mm_slot_cache)     /* initialization failed */
-               return NULL;
-       return kmem_cache_zalloc(mm_slot_cache, GFP_KERNEL);
+       return kmem_cache_alloc(stable_node_cache, GFP_KERNEL);
 }
 
-static inline void free_mm_slot(struct mm_slot *mm_slot)
+static inline void free_stable_node(struct stable_node *stable_node)
 {
-       kmem_cache_free(mm_slot_cache, mm_slot);
+       kmem_cache_free(stable_node_cache, stable_node);
 }
 
-static int __init mm_slots_hash_init(void)
+static inline struct mm_slot *alloc_mm_slot(void)
 {
-       mm_slots_hash = kzalloc(MM_SLOTS_HASH_HEADS * sizeof(struct hlist_head),
-                               GFP_KERNEL);
-       if (!mm_slots_hash)
-               return -ENOMEM;
-       return 0;
+       if (!mm_slot_cache)     /* initialization failed */
+               return NULL;
+       return kmem_cache_zalloc(mm_slot_cache, GFP_KERNEL);
 }
 
-static void __init mm_slots_hash_free(void)
+static inline void free_mm_slot(struct mm_slot *mm_slot)
 {
-       kfree(mm_slots_hash);
+       kmem_cache_free(mm_slot_cache, mm_slot);
 }
 
 static struct mm_slot *get_mm_slot(struct mm_struct *mm)
@@ -257,8 +279,7 @@ static struct mm_slot *get_mm_slot(struct mm_struct *mm)
        struct hlist_head *bucket;
        struct hlist_node *node;
 
-       bucket = &mm_slots_hash[((unsigned long)mm / sizeof(struct mm_struct))
-                               % MM_SLOTS_HASH_HEADS];
+       bucket = &mm_slots_hash[hash_ptr(mm, MM_SLOTS_HASH_SHIFT)];
        hlist_for_each_entry(mm_slot, node, bucket, link) {
                if (mm == mm_slot->mm)
                        return mm_slot;
@@ -271,10 +292,8 @@ static void insert_to_mm_slots_hash(struct mm_struct *mm,
 {
        struct hlist_head *bucket;
 
-       bucket = &mm_slots_hash[((unsigned long)mm / sizeof(struct mm_struct))
-                               % MM_SLOTS_HASH_HEADS];
+       bucket = &mm_slots_hash[hash_ptr(mm, MM_SLOTS_HASH_SHIFT)];
        mm_slot->mm = mm;
-       INIT_LIST_HEAD(&mm_slot->rmap_list);
        hlist_add_head(&mm_slot->link, bucket);
 }
 
@@ -284,6 +303,19 @@ static inline int in_stable_tree(struct rmap_item *rmap_item)
 }
 
 /*
+ * ksmd, and unmerge_and_remove_all_rmap_items(), must not touch an mm's
+ * page tables after it has passed through ksm_exit() - which, if necessary,
+ * takes mmap_sem briefly to serialize against them.  ksm_exit() does not set
+ * a special flag: they can just back out as soon as mm_users goes to zero.
+ * ksm_test_exit() is used throughout to make this test for exit: in some
+ * places for correctness, in some places just to avoid unnecessary work.
+ */
+static inline bool ksm_test_exit(struct mm_struct *mm)
+{
+       return atomic_read(&mm->mm_users) == 0;
+}
+
+/*
  * We use break_ksm to break COW on a ksm page: it's a stripped down
  *
  *     if (get_user_pages(current, mm, addr, 1, 1, 1, &page, NULL) == 1)
@@ -302,7 +334,7 @@ static int break_ksm(struct vm_area_struct *vma, unsigned long addr)
        do {
                cond_resched();
                page = follow_page(vma, addr, FOLL_GET);
-               if (!page)
+               if (IS_ERR_OR_NULL(page))
                        break;
                if (PageKsm(page))
                        ret = handle_mm_fault(vma->vm_mm, vma, addr,
@@ -342,21 +374,53 @@ static int break_ksm(struct vm_area_struct *vma, unsigned long addr)
        return (ret & VM_FAULT_OOM) ? -ENOMEM : 0;
 }
 
-static void break_cow(struct mm_struct *mm, unsigned long addr)
+static struct vm_area_struct *find_mergeable_vma(struct mm_struct *mm,
+               unsigned long addr)
 {
        struct vm_area_struct *vma;
-
-       down_read(&mm->mmap_sem);
+       if (ksm_test_exit(mm))
+               return NULL;
        vma = find_vma(mm, addr);
        if (!vma || vma->vm_start > addr)
-               goto out;
+               return NULL;
        if (!(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma)
-               goto out;
-       break_ksm(vma, addr);
-out:
+               return NULL;
+       return vma;
+}
+
+static void break_cow(struct rmap_item *rmap_item)
+{
+       struct mm_struct *mm = rmap_item->mm;
+       unsigned long addr = rmap_item->address;
+       struct vm_area_struct *vma;
+
+       /*
+        * It is not an accident that whenever we want to break COW
+        * to undo, we also need to drop a reference to the anon_vma.
+        */
+       put_anon_vma(rmap_item->anon_vma);
+
+       down_read(&mm->mmap_sem);
+       vma = find_mergeable_vma(mm, addr);
+       if (vma)
+               break_ksm(vma, addr);
        up_read(&mm->mmap_sem);
 }
 
+static struct page *page_trans_compound_anon(struct page *page)
+{
+       if (PageTransCompound(page)) {
+               struct page *head = compound_trans_head(page);
+               /*
+                * head may actually be splitted and freed from under
+                * us but it's ok here.
+                */
+               if (PageAnon(head))
+                       return head;
+       }
+       return NULL;
+}
+
 static struct page *get_mergeable_page(struct rmap_item *rmap_item)
 {
        struct mm_struct *mm = rmap_item->mm;
@@ -365,16 +429,14 @@ static struct page *get_mergeable_page(struct rmap_item *rmap_item)
        struct page *page;
 
        down_read(&mm->mmap_sem);
-       vma = find_vma(mm, addr);
-       if (!vma || vma->vm_start > addr)
-               goto out;
-       if (!(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma)
+       vma = find_mergeable_vma(mm, addr);
+       if (!vma)
                goto out;
 
        page = follow_page(vma, addr, FOLL_GET);
-       if (!page)
+       if (IS_ERR_OR_NULL(page))
                goto out;
-       if (PageAnon(page)) {
+       if (PageAnon(page) || page_trans_compound_anon(page)) {
                flush_anon_page(vma, page, addr);
                flush_dcache_page(page);
        } else {
@@ -385,21 +447,77 @@ out:              page = NULL;
        return page;
 }
 
+static void remove_node_from_stable_tree(struct stable_node *stable_node)
+{
+       struct rmap_item *rmap_item;
+       struct hlist_node *hlist;
+
+       hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) {
+               if (rmap_item->hlist.next)
+                       ksm_pages_sharing--;
+               else
+                       ksm_pages_shared--;
+               put_anon_vma(rmap_item->anon_vma);
+               rmap_item->address &= PAGE_MASK;
+               cond_resched();
+       }
+
+       rb_erase(&stable_node->node, &root_stable_tree);
+       free_stable_node(stable_node);
+}
+
 /*
- * get_ksm_page: checks if the page at the virtual address in rmap_item
- * is still PageKsm, in which case we can trust the content of the page,
- * and it returns the gotten page; but NULL if the page has been zapped.
+ * get_ksm_page: checks if the page indicated by the stable node
+ * is still its ksm page, despite having held no reference to it.
+ * In which case we can trust the content of the page, and it
+ * returns the gotten page; but if the page has now been zapped,
+ * remove the stale node from the stable tree and return NULL.
+ *
+ * You would expect the stable_node to hold a reference to the ksm page.
+ * But if it increments the page's count, swapping out has to wait for
+ * ksmd to come around again before it can free the page, which may take
+ * seconds or even minutes: much too unresponsive.  So instead we use a
+ * "keyhole reference": access to the ksm page from the stable node peeps
+ * out through its keyhole to see if that page still holds the right key,
+ * pointing back to this stable node.  This relies on freeing a PageAnon
+ * page to reset its page->mapping to NULL, and relies on no other use of
+ * a page to put something that might look like our key in page->mapping.
+ *
+ * include/linux/pagemap.h page_cache_get_speculative() is a good reference,
+ * but this is different - made simpler by ksm_thread_mutex being held, but
+ * interesting for assuming that no other use of the struct page could ever
+ * put our expected_mapping into page->mapping (or a field of the union which
+ * coincides with page->mapping).  The RCU calls are not for KSM at all, but
+ * to keep the page_count protocol described with page_cache_get_speculative.
+ *
+ * Note: it is possible that get_ksm_page() will return NULL one moment,
+ * then page the next, if the page is in between page_freeze_refs() and
+ * page_unfreeze_refs(): this shouldn't be a problem anywhere, the page
+ * is on its way to being freed; but it is an anomaly to bear in mind.
  */
-static struct page *get_ksm_page(struct rmap_item *rmap_item)
+static struct page *get_ksm_page(struct stable_node *stable_node)
 {
        struct page *page;
-
-       page = get_mergeable_page(rmap_item);
-       if (page && !PageKsm(page)) {
+       void *expected_mapping;
+
+       page = pfn_to_page(stable_node->kpfn);
+       expected_mapping = (void *)stable_node +
+                               (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM);
+       rcu_read_lock();
+       if (page->mapping != expected_mapping)
+               goto stale;
+       if (!get_page_unless_zero(page))
+               goto stale;
+       if (page->mapping != expected_mapping) {
                put_page(page);
-               page = NULL;
+               goto stale;
        }
+       rcu_read_unlock();
        return page;
+stale:
+       rcu_read_unlock();
+       remove_node_from_stable_tree(stable_node);
+       return NULL;
 }
 
 /*
@@ -408,70 +526,56 @@ static struct page *get_ksm_page(struct rmap_item *rmap_item)
  */
 static void remove_rmap_item_from_tree(struct rmap_item *rmap_item)
 {
-       if (in_stable_tree(rmap_item)) {
-               struct rmap_item *next_item = rmap_item->next;
-
-               if (rmap_item->address & NODE_FLAG) {
-                       if (next_item) {
-                               rb_replace_node(&rmap_item->node,
-                                               &next_item->node,
-                                               &root_stable_tree);
-                               next_item->address |= NODE_FLAG;
-                               ksm_pages_sharing--;
-                       } else {
-                               rb_erase(&rmap_item->node, &root_stable_tree);
-                               ksm_pages_shared--;
-                       }
-               } else {
-                       struct rmap_item *prev_item = rmap_item->prev;
+       if (rmap_item->address & STABLE_FLAG) {
+               struct stable_node *stable_node;
+               struct page *page;
 
-                       BUG_ON(prev_item->next != rmap_item);
-                       prev_item->next = next_item;
-                       if (next_item) {
-                               BUG_ON(next_item->prev != rmap_item);
-                               next_item->prev = rmap_item->prev;
-                       }
+               stable_node = rmap_item->head;
+               page = get_ksm_page(stable_node);
+               if (!page)
+                       goto out;
+
+               lock_page(page);
+               hlist_del(&rmap_item->hlist);
+               unlock_page(page);
+               put_page(page);
+
+               if (stable_node->hlist.first)
                        ksm_pages_sharing--;
-               }
+               else
+                       ksm_pages_shared--;
 
-               rmap_item->next = NULL;
+               put_anon_vma(rmap_item->anon_vma);
+               rmap_item->address &= PAGE_MASK;
 
-       } else if (rmap_item->address & NODE_FLAG) {
+       } else if (rmap_item->address & UNSTABLE_FLAG) {
                unsigned char age;
                /*
-                * ksm_thread can and must skip the rb_erase, because
+                * Usually ksmd can and must skip the rb_erase, because
                 * root_unstable_tree was already reset to RB_ROOT.
-                * But __ksm_exit has to be careful: do the rb_erase
-                * if it's interrupting a scan, and this rmap_item was
-                * inserted by this scan rather than left from before.
-                *
-                * Because of the case in which remove_mm_from_lists
-                * increments seqnr before removing rmaps, unstable_nr
-                * may even be 2 behind seqnr, but should never be
-                * further behind.  Yes, I did have trouble with this!
+                * But be careful when an mm is exiting: do the rb_erase
+                * if this rmap_item was inserted by this scan, rather
+                * than left over from before.
                 */
                age = (unsigned char)(ksm_scan.seqnr - rmap_item->address);
-               BUG_ON(age > 2);
+               BUG_ON(age > 1);
                if (!age)
                        rb_erase(&rmap_item->node, &root_unstable_tree);
+
                ksm_pages_unshared--;
+               rmap_item->address &= PAGE_MASK;
        }
-
-       rmap_item->address &= PAGE_MASK;
-
+out:
        cond_resched();         /* we're called from many long loops */
 }
 
 static void remove_trailing_rmap_items(struct mm_slot *mm_slot,
-                                      struct list_head *cur)
+                                      struct rmap_item **rmap_list)
 {
-       struct rmap_item *rmap_item;
-
-       while (cur != &mm_slot->rmap_list) {
-               rmap_item = list_entry(cur, struct rmap_item, link);
-               cur = cur->next;
+       while (*rmap_list) {
+               struct rmap_item *rmap_item = *rmap_list;
+               *rmap_list = rmap_item->rmap_list;
                remove_rmap_item_from_tree(rmap_item);
-               list_del(&rmap_item->link);
                free_rmap_item(rmap_item);
        }
 }
@@ -496,6 +600,8 @@ static int unmerge_ksm_pages(struct vm_area_struct *vma,
        int err = 0;
 
        for (addr = start; addr < end && !err; addr += PAGE_SIZE) {
+               if (ksm_test_exit(vma->vm_mm))
+                       break;
                if (signal_pending(current))
                        err = -ERESTARTSYS;
                else
@@ -504,6 +610,10 @@ static int unmerge_ksm_pages(struct vm_area_struct *vma,
        return err;
 }
 
+#ifdef CONFIG_SYSFS
+/*
+ * Only called through the sysfs control interface:
+ */
 static int unmerge_and_remove_all_rmap_items(void)
 {
        struct mm_slot *mm_slot;
@@ -512,77 +622,63 @@ static int unmerge_and_remove_all_rmap_items(void)
        int err = 0;
 
        spin_lock(&ksm_mmlist_lock);
-       mm_slot = list_entry(ksm_mm_head.mm_list.next,
+       ksm_scan.mm_slot = list_entry(ksm_mm_head.mm_list.next,
                                                struct mm_slot, mm_list);
        spin_unlock(&ksm_mmlist_lock);
 
-       while (mm_slot != &ksm_mm_head) {
+       for (mm_slot = ksm_scan.mm_slot;
+                       mm_slot != &ksm_mm_head; mm_slot = ksm_scan.mm_slot) {
                mm = mm_slot->mm;
                down_read(&mm->mmap_sem);
                for (vma = mm->mmap; vma; vma = vma->vm_next) {
+                       if (ksm_test_exit(mm))
+                               break;
                        if (!(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma)
                                continue;
                        err = unmerge_ksm_pages(vma,
                                                vma->vm_start, vma->vm_end);
-                       if (err) {
-                               up_read(&mm->mmap_sem);
-                               goto out;
-                       }
+                       if (err)
+                               goto error;
                }
-               remove_trailing_rmap_items(mm_slot, mm_slot->rmap_list.next);
-               up_read(&mm->mmap_sem);
+
+               remove_trailing_rmap_items(mm_slot, &mm_slot->rmap_list);
 
                spin_lock(&ksm_mmlist_lock);
-               mm_slot = list_entry(mm_slot->mm_list.next,
+               ksm_scan.mm_slot = list_entry(mm_slot->mm_list.next,
                                                struct mm_slot, mm_list);
-               spin_unlock(&ksm_mmlist_lock);
+               if (ksm_test_exit(mm)) {
+                       hlist_del(&mm_slot->link);
+                       list_del(&mm_slot->mm_list);
+                       spin_unlock(&ksm_mmlist_lock);
+
+                       free_mm_slot(mm_slot);
+                       clear_bit(MMF_VM_MERGEABLE, &mm->flags);
+                       up_read(&mm->mmap_sem);
+                       mmdrop(mm);
+               } else {
+                       spin_unlock(&ksm_mmlist_lock);
+                       up_read(&mm->mmap_sem);
+               }
        }
 
        ksm_scan.seqnr = 0;
-out:
+       return 0;
+
+error:
+       up_read(&mm->mmap_sem);
        spin_lock(&ksm_mmlist_lock);
        ksm_scan.mm_slot = &ksm_mm_head;
        spin_unlock(&ksm_mmlist_lock);
        return err;
 }
-
-static void remove_mm_from_lists(struct mm_struct *mm)
-{
-       struct mm_slot *mm_slot;
-
-       spin_lock(&ksm_mmlist_lock);
-       mm_slot = get_mm_slot(mm);
-
-       /*
-        * This mm_slot is always at the scanning cursor when we're
-        * called from scan_get_next_rmap_item; but it's a special
-        * case when we're called from __ksm_exit.
-        */
-       if (ksm_scan.mm_slot == mm_slot) {
-               ksm_scan.mm_slot = list_entry(
-                       mm_slot->mm_list.next, struct mm_slot, mm_list);
-               ksm_scan.address = 0;
-               ksm_scan.rmap_item = list_entry(
-                       &ksm_scan.mm_slot->rmap_list, struct rmap_item, link);
-               if (ksm_scan.mm_slot == &ksm_mm_head)
-                       ksm_scan.seqnr++;
-       }
-
-       hlist_del(&mm_slot->link);
-       list_del(&mm_slot->mm_list);
-       spin_unlock(&ksm_mmlist_lock);
-
-       remove_trailing_rmap_items(mm_slot, mm_slot->rmap_list.next);
-       free_mm_slot(mm_slot);
-       clear_bit(MMF_VM_MERGEABLE, &mm->flags);
-}
+#endif /* CONFIG_SYSFS */
 
 static u32 calc_checksum(struct page *page)
 {
        u32 checksum;
-       void *addr = kmap_atomic(page, KM_USER0);
+       void *addr = kmap_atomic(page);
        checksum = jhash2(addr, PAGE_SIZE / 4, 17);
-       kunmap_atomic(addr, KM_USER0);
+       kunmap_atomic(addr);
        return checksum;
 }
 
@@ -591,11 +687,11 @@ static int memcmp_pages(struct page *page1, struct page *page2)
        char *addr1, *addr2;
        int ret;
 
-       addr1 = kmap_atomic(page1, KM_USER0);
-       addr2 = kmap_atomic(page2, KM_USER1);
+       addr1 = kmap_atomic(page1);
+       addr2 = kmap_atomic(page2);
        ret = memcmp(addr1, addr2, PAGE_SIZE);
-       kunmap_atomic(addr2, KM_USER1);
-       kunmap_atomic(addr1, KM_USER0);
+       kunmap_atomic(addr2);
+       kunmap_atomic(addr1);
        return ret;
 }
 
@@ -613,22 +709,30 @@ static int write_protect_page(struct vm_area_struct *vma, struct page *page,
        spinlock_t *ptl;
        int swapped;
        int err = -EFAULT;
+       unsigned long mmun_start;       /* For mmu_notifiers */
+       unsigned long mmun_end;         /* For mmu_notifiers */
 
        addr = page_address_in_vma(page, vma);
        if (addr == -EFAULT)
                goto out;
 
+       BUG_ON(PageTransCompound(page));
+
+       mmun_start = addr;
+       mmun_end   = addr + PAGE_SIZE;
+       mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
+
        ptep = page_check_address(page, mm, addr, &ptl, 0);
        if (!ptep)
-               goto out;
+               goto out_mn;
 
-       if (pte_write(*ptep)) {
+       if (pte_write(*ptep) || pte_dirty(*ptep)) {
                pte_t entry;
 
                swapped = PageSwapCache(page);
                flush_cache_page(vma, addr, page_to_pfn(page));
                /*
-                * Ok this is tricky, when get_user_pages_fast() run it doesnt
+                * Ok this is tricky, when get_user_pages_fast() run it doesn't
                 * take any lock, therefore the check that we are going to make
                 * with the pagecount against the mapcount is racey and
                 * O_DIRECT can happen right after the check.
@@ -641,11 +745,13 @@ static int write_protect_page(struct vm_area_struct *vma, struct page *page,
                 * Check that no O_DIRECT or similar I/O is in progress on the
                 * page
                 */
-               if ((page_mapcount(page) + 2 + swapped) != page_count(page)) {
-                       set_pte_at_notify(mm, addr, ptep, entry);
+               if (page_mapcount(page) + 1 + swapped != page_count(page)) {
+                       set_pte_at(mm, addr, ptep, entry);
                        goto out_unlock;
                }
-               entry = pte_wrprotect(entry);
+               if (pte_dirty(entry))
+                       set_page_dirty(page);
+               entry = pte_mkclean(pte_wrprotect(entry));
                set_pte_at_notify(mm, addr, ptep, entry);
        }
        *orig_pte = *ptep;
@@ -653,99 +759,124 @@ static int write_protect_page(struct vm_area_struct *vma, struct page *page,
 
 out_unlock:
        pte_unmap_unlock(ptep, ptl);
+out_mn:
+       mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
 out:
        return err;
 }
 
 /**
  * replace_page - replace page in vma by new ksm page
- * @vma:      vma that holds the pte pointing to oldpage
- * @oldpage:  the page we are replacing by newpage
- * @newpage:  the ksm page we replace oldpage by
+ * @vma:      vma that holds the pte pointing to page
+ * @page:     the page we are replacing by kpage
+ * @kpage:    the ksm page we replace page by
  * @orig_pte: the original value of the pte
  *
  * Returns 0 on success, -EFAULT on failure.
  */
-static int replace_page(struct vm_area_struct *vma, struct page *oldpage,
-                       struct page *newpage, pte_t orig_pte)
+static int replace_page(struct vm_area_struct *vma, struct page *page,
+                       struct page *kpage, pte_t orig_pte)
 {
        struct mm_struct *mm = vma->vm_mm;
-       pgd_t *pgd;
-       pud_t *pud;
        pmd_t *pmd;
        pte_t *ptep;
        spinlock_t *ptl;
        unsigned long addr;
-       pgprot_t prot;
        int err = -EFAULT;
+       unsigned long mmun_start;       /* For mmu_notifiers */
+       unsigned long mmun_end;         /* For mmu_notifiers */
 
-       prot = vm_get_page_prot(vma->vm_flags & ~VM_WRITE);
-
-       addr = page_address_in_vma(oldpage, vma);
+       addr = page_address_in_vma(page, vma);
        if (addr == -EFAULT)
                goto out;
 
-       pgd = pgd_offset(mm, addr);
-       if (!pgd_present(*pgd))
+       pmd = mm_find_pmd(mm, addr);
+       if (!pmd)
                goto out;
+       BUG_ON(pmd_trans_huge(*pmd));
 
-       pud = pud_offset(pgd, addr);
-       if (!pud_present(*pud))
-               goto out;
-
-       pmd = pmd_offset(pud, addr);
-       if (!pmd_present(*pmd))
-               goto out;
+       mmun_start = addr;
+       mmun_end   = addr + PAGE_SIZE;
+       mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
 
        ptep = pte_offset_map_lock(mm, pmd, addr, &ptl);
        if (!pte_same(*ptep, orig_pte)) {
                pte_unmap_unlock(ptep, ptl);
-               goto out;
+               goto out_mn;
        }
 
-       get_page(newpage);
-       page_add_ksm_rmap(newpage);
+       get_page(kpage);
+       page_add_anon_rmap(kpage, vma, addr);
 
        flush_cache_page(vma, addr, pte_pfn(*ptep));
        ptep_clear_flush(vma, addr, ptep);
-       set_pte_at_notify(mm, addr, ptep, mk_pte(newpage, prot));
+       set_pte_at_notify(mm, addr, ptep, mk_pte(kpage, vma->vm_page_prot));
 
-       page_remove_rmap(oldpage);
-       put_page(oldpage);
+       page_remove_rmap(page);
+       if (!page_mapped(page))
+               try_to_free_swap(page);
+       put_page(page);
 
        pte_unmap_unlock(ptep, ptl);
        err = 0;
+out_mn:
+       mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
 out:
        return err;
 }
 
+static int page_trans_compound_anon_split(struct page *page)
+{
+       int ret = 0;
+       struct page *transhuge_head = page_trans_compound_anon(page);
+       if (transhuge_head) {
+               /* Get the reference on the head to split it. */
+               if (get_page_unless_zero(transhuge_head)) {
+                       /*
+                        * Recheck we got the reference while the head
+                        * was still anonymous.
+                        */
+                       if (PageAnon(transhuge_head))
+                               ret = split_huge_page(transhuge_head);
+                       else
+                               /*
+                                * Retry later if split_huge_page run
+                                * from under us.
+                                */
+                               ret = 1;
+                       put_page(transhuge_head);
+               } else
+                       /* Retry later if split_huge_page run from under us. */
+                       ret = 1;
+       }
+       return ret;
+}
+
 /*
  * try_to_merge_one_page - take two pages and merge them into one
- * @vma: the vma that hold the pte pointing into oldpage
- * @oldpage: the page that we want to replace with newpage
- * @newpage: the page that we want to map instead of oldpage
- *
- * Note:
- * oldpage should be a PageAnon page, while newpage should be a PageKsm page,
- * or a newly allocated kernel page which page_add_ksm_rmap will make PageKsm.
+ * @vma: the vma that holds the pte pointing to page
+ * @page: the PageAnon page that we want to replace with kpage
+ * @kpage: the PageKsm page that we want to map instead of page,
+ *         or NULL the first time when we want to use page as kpage.
  *
  * This function returns 0 if the pages were merged, -EFAULT otherwise.
  */
 static int try_to_merge_one_page(struct vm_area_struct *vma,
-                                struct page *oldpage,
-                                struct page *newpage)
+                                struct page *page, struct page *kpage)
 {
        pte_t orig_pte = __pte(0);
        int err = -EFAULT;
 
+       if (page == kpage)                      /* ksm page forked */
+               return 0;
+
        if (!(vma->vm_flags & VM_MERGEABLE))
                goto out;
-
-       if (!PageAnon(oldpage))
+       if (PageTransCompound(page) && page_trans_compound_anon_split(page))
+               goto out;
+       BUG_ON(PageTransCompound(page));
+       if (!PageAnon(page))
                goto out;
-
-       get_page(newpage);
-       get_page(oldpage);
 
        /*
         * We need the page lock to read a stable PageSwapCache in
@@ -754,26 +885,39 @@ static int try_to_merge_one_page(struct vm_area_struct *vma,
         * prefer to continue scanning and merging different pages,
         * then come back to this page when it is unlocked.
         */
-       if (!trylock_page(oldpage))
-               goto out_putpage;
+       if (!trylock_page(page))
+               goto out;
        /*
         * If this anonymous page is mapped only here, its pte may need
         * to be write-protected.  If it's mapped elsewhere, all of its
         * ptes are necessarily already write-protected.  But in either
         * case, we need to lock and check page_count is not raised.
         */
-       if (write_protect_page(vma, oldpage, &orig_pte)) {
-               unlock_page(oldpage);
-               goto out_putpage;
+       if (write_protect_page(vma, page, &orig_pte) == 0) {
+               if (!kpage) {
+                       /*
+                        * While we hold page lock, upgrade page from
+                        * PageAnon+anon_vma to PageKsm+NULL stable_node:
+                        * stable_tree_insert() will update stable_node.
+                        */
+                       set_page_stable_node(page, NULL);
+                       mark_page_accessed(page);
+                       err = 0;
+               } else if (pages_identical(page, kpage))
+                       err = replace_page(vma, page, kpage, orig_pte);
        }
-       unlock_page(oldpage);
 
-       if (pages_identical(oldpage, newpage))
-               err = replace_page(vma, oldpage, newpage, orig_pte);
+       if ((vma->vm_flags & VM_LOCKED) && kpage && !err) {
+               munlock_vma_page(page);
+               if (!PageMlocked(kpage)) {
+                       unlock_page(page);
+                       lock_page(kpage);
+                       mlock_vma_page(kpage);
+                       page = kpage;           /* for final unlock */
+               }
+       }
 
-out_putpage:
-       put_page(oldpage);
-       put_page(newpage);
+       unlock_page(page);
 out:
        return err;
 }
@@ -781,23 +925,32 @@ out:
 /*
  * try_to_merge_with_ksm_page - like try_to_merge_two_pages,
  * but no new kernel page is allocated: kpage must already be a ksm page.
+ *
+ * This function returns 0 if the pages were merged, -EFAULT otherwise.
  */
-static int try_to_merge_with_ksm_page(struct mm_struct *mm1,
-                                     unsigned long addr1,
-                                     struct page *page1,
-                                     struct page *kpage)
+static int try_to_merge_with_ksm_page(struct rmap_item *rmap_item,
+                                     struct page *page, struct page *kpage)
 {
+       struct mm_struct *mm = rmap_item->mm;
        struct vm_area_struct *vma;
        int err = -EFAULT;
 
-       down_read(&mm1->mmap_sem);
-       vma = find_vma(mm1, addr1);
-       if (!vma || vma->vm_start > addr1)
+       down_read(&mm->mmap_sem);
+       if (ksm_test_exit(mm))
+               goto out;
+       vma = find_vma(mm, rmap_item->address);
+       if (!vma || vma->vm_start > rmap_item->address)
                goto out;
 
-       err = try_to_merge_one_page(vma, page1, kpage);
+       err = try_to_merge_one_page(vma, page, kpage);
+       if (err)
+               goto out;
+
+       /* Must get reference to anon_vma while still holding mmap_sem */
+       rmap_item->anon_vma = vma->anon_vma;
+       get_anon_vma(vma->anon_vma);
 out:
-       up_read(&mm1->mmap_sem);
+       up_read(&mm->mmap_sem);
        return err;
 }
 
@@ -805,105 +958,73 @@ out:
  * try_to_merge_two_pages - take two identical pages and prepare them
  * to be merged into one page.
  *
- * This function returns 0 if we successfully mapped two identical pages
- * into one page, -EFAULT otherwise.
+ * This function returns the kpage if we successfully merged two identical
+ * pages into one ksm page, NULL otherwise.
  *
- * Note that this function allocates a new kernel page: if one of the pages
+ * Note that this function upgrades page to ksm page: if one of the pages
  * is already a ksm page, try_to_merge_with_ksm_page should be used.
  */
-static int try_to_merge_two_pages(struct mm_struct *mm1, unsigned long addr1,
-                                 struct page *page1, struct mm_struct *mm2,
-                                 unsigned long addr2, struct page *page2)
+static struct page *try_to_merge_two_pages(struct rmap_item *rmap_item,
+                                          struct page *page,
+                                          struct rmap_item *tree_rmap_item,
+                                          struct page *tree_page)
 {
-       struct vm_area_struct *vma;
-       struct page *kpage;
-       int err = -EFAULT;
-
-       /*
-        * The number of nodes in the stable tree
-        * is the number of kernel pages that we hold.
-        */
-       if (ksm_max_kernel_pages &&
-           ksm_max_kernel_pages <= ksm_pages_shared)
-               return err;
-
-       kpage = alloc_page(GFP_HIGHUSER);
-       if (!kpage)
-               return err;
-
-       down_read(&mm1->mmap_sem);
-       vma = find_vma(mm1, addr1);
-       if (!vma || vma->vm_start > addr1) {
-               up_read(&mm1->mmap_sem);
-               goto out;
-       }
-
-       copy_user_highpage(kpage, page1, addr1, vma);
-       err = try_to_merge_one_page(vma, page1, kpage);
-       up_read(&mm1->mmap_sem);
+       int err;
 
+       err = try_to_merge_with_ksm_page(rmap_item, page, NULL);
        if (!err) {
-               err = try_to_merge_with_ksm_page(mm2, addr2, page2, kpage);
+               err = try_to_merge_with_ksm_page(tree_rmap_item,
+                                                       tree_page, page);
                /*
                 * If that fails, we have a ksm page with only one pte
                 * pointing to it: so break it.
                 */
                if (err)
-                       break_cow(mm1, addr1);
+                       break_cow(rmap_item);
        }
-out:
-       put_page(kpage);
-       return err;
+       return err ? NULL : page;
 }
 
 /*
- * stable_tree_search - search page inside the stable tree
- * @page: the page that we are searching identical pages to.
- * @page2: pointer into identical page that we are holding inside the stable
- *        tree that we have found.
- * @rmap_item: the reverse mapping item
+ * stable_tree_search - search for page inside the stable tree
  *
  * This function checks if there is a page inside the stable tree
  * with identical content to the page that we are scanning right now.
  *
- * This function return rmap_item pointer to the identical item if found,
+ * This function returns the stable tree node of identical content if found,
  * NULL otherwise.
  */
-static struct rmap_item *stable_tree_search(struct page *page,
-                                           struct page **page2,
-                                           struct rmap_item *rmap_item)
+static struct page *stable_tree_search(struct page *page)
 {
        struct rb_node *node = root_stable_tree.rb_node;
+       struct stable_node *stable_node;
+
+       stable_node = page_stable_node(page);
+       if (stable_node) {                      /* ksm page forked */
+               get_page(page);
+               return page;
+       }
 
        while (node) {
-               struct rmap_item *tree_rmap_item, *next_rmap_item;
+               struct page *tree_page;
                int ret;
 
-               tree_rmap_item = rb_entry(node, struct rmap_item, node);
-               while (tree_rmap_item) {
-                       BUG_ON(!in_stable_tree(tree_rmap_item));
-                       cond_resched();
-                       page2[0] = get_ksm_page(tree_rmap_item);
-                       if (page2[0])
-                               break;
-                       next_rmap_item = tree_rmap_item->next;
-                       remove_rmap_item_from_tree(tree_rmap_item);
-                       tree_rmap_item = next_rmap_item;
-               }
-               if (!tree_rmap_item)
+               cond_resched();
+               stable_node = rb_entry(node, struct stable_node, node);
+               tree_page = get_ksm_page(stable_node);
+               if (!tree_page)
                        return NULL;
 
-               ret = memcmp_pages(page, page2[0]);
+               ret = memcmp_pages(page, tree_page);
 
                if (ret < 0) {
-                       put_page(page2[0]);
+                       put_page(tree_page);
                        node = node->rb_left;
                } else if (ret > 0) {
-                       put_page(page2[0]);
+                       put_page(tree_page);
                        node = node->rb_right;
-               } else {
-                       return tree_rmap_item;
-               }
+               } else
+                       return tree_page;
        }
 
        return NULL;
@@ -913,38 +1034,26 @@ static struct rmap_item *stable_tree_search(struct page *page,
  * stable_tree_insert - insert rmap_item pointing to new ksm page
  * into the stable tree.
  *
- * @page: the page that we are searching identical page to inside the stable
- *       tree.
- * @rmap_item: pointer to the reverse mapping item.
- *
- * This function returns rmap_item if success, NULL otherwise.
+ * This function returns the stable tree node just allocated on success,
+ * NULL otherwise.
  */
-static struct rmap_item *stable_tree_insert(struct page *page,
-                                           struct rmap_item *rmap_item)
+static struct stable_node *stable_tree_insert(struct page *kpage)
 {
        struct rb_node **new = &root_stable_tree.rb_node;
        struct rb_node *parent = NULL;
+       struct stable_node *stable_node;
 
        while (*new) {
-               struct rmap_item *tree_rmap_item, *next_rmap_item;
                struct page *tree_page;
                int ret;
 
-               tree_rmap_item = rb_entry(*new, struct rmap_item, node);
-               while (tree_rmap_item) {
-                       BUG_ON(!in_stable_tree(tree_rmap_item));
-                       cond_resched();
-                       tree_page = get_ksm_page(tree_rmap_item);
-                       if (tree_page)
-                               break;
-                       next_rmap_item = tree_rmap_item->next;
-                       remove_rmap_item_from_tree(tree_rmap_item);
-                       tree_rmap_item = next_rmap_item;
-               }
-               if (!tree_rmap_item)
+               cond_resched();
+               stable_node = rb_entry(*new, struct stable_node, node);
+               tree_page = get_ksm_page(stable_node);
+               if (!tree_page)
                        return NULL;
 
-               ret = memcmp_pages(page, tree_page);
+               ret = memcmp_pages(kpage, tree_page);
                put_page(tree_page);
 
                parent = *new;
@@ -962,22 +1071,24 @@ static struct rmap_item *stable_tree_insert(struct page *page,
                }
        }
 
-       rmap_item->address |= NODE_FLAG | STABLE_FLAG;
-       rmap_item->next = NULL;
-       rb_link_node(&rmap_item->node, parent, new);
-       rb_insert_color(&rmap_item->node, &root_stable_tree);
+       stable_node = alloc_stable_node();
+       if (!stable_node)
+               return NULL;
 
-       ksm_pages_shared++;
-       return rmap_item;
+       rb_link_node(&stable_node->node, parent, new);
+       rb_insert_color(&stable_node->node, &root_stable_tree);
+
+       INIT_HLIST_HEAD(&stable_node->hlist);
+
+       stable_node->kpfn = page_to_pfn(kpage);
+       set_page_stable_node(kpage, stable_node);
+
+       return stable_node;
 }
 
 /*
- * unstable_tree_search_insert - search and insert items into the unstable tree.
- *
- * @page: the page that we are going to search for identical page or to insert
- *       into the unstable tree
- * @page2: pointer into identical page that was found inside the unstable tree
- * @rmap_item: the reverse mapping item of page
+ * unstable_tree_search_insert - search for identical page,
+ * else insert rmap_item into the unstable tree.
  *
  * This function searches for a page in the unstable tree identical to the
  * page currently being scanned; and if no identical page is found in the
@@ -989,46 +1100,50 @@ static struct rmap_item *stable_tree_insert(struct page *page,
  * This function does both searching and inserting, because they share
  * the same walking algorithm in an rbtree.
  */
-static struct rmap_item *unstable_tree_search_insert(struct page *page,
-                                               struct page **page2,
-                                               struct rmap_item *rmap_item)
+static
+struct rmap_item *unstable_tree_search_insert(struct rmap_item *rmap_item,
+                                             struct page *page,
+                                             struct page **tree_pagep)
+
 {
        struct rb_node **new = &root_unstable_tree.rb_node;
        struct rb_node *parent = NULL;
 
        while (*new) {
                struct rmap_item *tree_rmap_item;
+               struct page *tree_page;
                int ret;
 
+               cond_resched();
                tree_rmap_item = rb_entry(*new, struct rmap_item, node);
-               page2[0] = get_mergeable_page(tree_rmap_item);
-               if (!page2[0])
+               tree_page = get_mergeable_page(tree_rmap_item);
+               if (IS_ERR_OR_NULL(tree_page))
                        return NULL;
 
                /*
-                * Don't substitute an unswappable ksm page
-                * just for one good swappable forked page.
+                * Don't substitute a ksm page for a forked page.
                 */
-               if (page == page2[0]) {
-                       put_page(page2[0]);
+               if (page == tree_page) {
+                       put_page(tree_page);
                        return NULL;
                }
 
-               ret = memcmp_pages(page, page2[0]);
+               ret = memcmp_pages(page, tree_page);
 
                parent = *new;
                if (ret < 0) {
-                       put_page(page2[0]);
+                       put_page(tree_page);
                        new = &parent->rb_left;
                } else if (ret > 0) {
-                       put_page(page2[0]);
+                       put_page(tree_page);
                        new = &parent->rb_right;
                } else {
+                       *tree_pagep = tree_page;
                        return tree_rmap_item;
                }
        }
 
-       rmap_item->address |= NODE_FLAG;
+       rmap_item->address |= UNSTABLE_FLAG;
        rmap_item->address |= (ksm_scan.seqnr & SEQNR_MASK);
        rb_link_node(&rmap_item->node, parent, new);
        rb_insert_color(&rmap_item->node, &root_unstable_tree);
@@ -1043,18 +1158,16 @@ static struct rmap_item *unstable_tree_search_insert(struct page *page,
  * the same ksm page.
  */
 static void stable_tree_append(struct rmap_item *rmap_item,
-                              struct rmap_item *tree_rmap_item)
+                              struct stable_node *stable_node)
 {
-       rmap_item->next = tree_rmap_item->next;
-       rmap_item->prev = tree_rmap_item;
-
-       if (tree_rmap_item->next)
-               tree_rmap_item->next->prev = rmap_item;
-
-       tree_rmap_item->next = rmap_item;
+       rmap_item->head = stable_node;
        rmap_item->address |= STABLE_FLAG;
+       hlist_add_head(&rmap_item->hlist, &stable_node->hlist);
 
-       ksm_pages_sharing++;
+       if (rmap_item->hlist.next)
+               ksm_pages_sharing++;
+       else
+               ksm_pages_shared++;
 }
 
 /*
@@ -1068,49 +1181,37 @@ static void stable_tree_append(struct rmap_item *rmap_item,
  */
 static void cmp_and_merge_page(struct page *page, struct rmap_item *rmap_item)
 {
-       struct page *page2[1];
        struct rmap_item *tree_rmap_item;
+       struct page *tree_page = NULL;
+       struct stable_node *stable_node;
+       struct page *kpage;
        unsigned int checksum;
        int err;
 
-       if (in_stable_tree(rmap_item))
-               remove_rmap_item_from_tree(rmap_item);
+       remove_rmap_item_from_tree(rmap_item);
 
        /* We first start with searching the page inside the stable tree */
-       tree_rmap_item = stable_tree_search(page, page2, rmap_item);
-       if (tree_rmap_item) {
-               if (page == page2[0])                   /* forked */
-                       err = 0;
-               else
-                       err = try_to_merge_with_ksm_page(rmap_item->mm,
-                                                        rmap_item->address,
-                                                        page, page2[0]);
-               put_page(page2[0]);
-
+       kpage = stable_tree_search(page);
+       if (kpage) {
+               err = try_to_merge_with_ksm_page(rmap_item, page, kpage);
                if (!err) {
                        /*
                         * The page was successfully merged:
                         * add its rmap_item to the stable tree.
                         */
-                       stable_tree_append(rmap_item, tree_rmap_item);
+                       lock_page(kpage);
+                       stable_tree_append(rmap_item, page_stable_node(kpage));
+                       unlock_page(kpage);
                }
+               put_page(kpage);
                return;
        }
 
        /*
-        * A ksm page might have got here by fork, but its other
-        * references have already been removed from the stable tree.
-        * Or it might be left over from a break_ksm which failed
-        * when the mem_cgroup had reached its limit: try again now.
-        */
-       if (PageKsm(page))
-               break_cow(rmap_item->mm, rmap_item->address);
-
-       /*
-        * In case the hash value of the page was changed from the last time we
-        * have calculated it, this page to be changed frequely, therefore we
-        * don't want to insert it to the unstable tree, and we don't want to
-        * waste our time to search if there is something identical to it there.
+        * If the hash value of the page has changed from the last time
+        * we calculated it, this page is changing frequently: therefore we
+        * don't want to insert it in the unstable tree, and we don't want
+        * to waste our time searching for something identical to it there.
         */
        checksum = calc_checksum(page);
        if (rmap_item->oldchecksum != checksum) {
@@ -1118,21 +1219,27 @@ static void cmp_and_merge_page(struct page *page, struct rmap_item *rmap_item)
                return;
        }
 
-       tree_rmap_item = unstable_tree_search_insert(page, page2, rmap_item);
+       tree_rmap_item =
+               unstable_tree_search_insert(rmap_item, page, &tree_page);
        if (tree_rmap_item) {
-               err = try_to_merge_two_pages(rmap_item->mm,
-                                            rmap_item->address, page,
-                                            tree_rmap_item->mm,
-                                            tree_rmap_item->address, page2[0]);
+               kpage = try_to_merge_two_pages(rmap_item, page,
+                                               tree_rmap_item, tree_page);
+               put_page(tree_page);
                /*
                 * As soon as we merge this page, we want to remove the
                 * rmap_item of the page we have merged with from the unstable
                 * tree, and insert it instead as new node in the stable tree.
                 */
-               if (!err) {
-                       rb_erase(&tree_rmap_item->node, &root_unstable_tree);
-                       tree_rmap_item->address &= ~NODE_FLAG;
-                       ksm_pages_unshared--;
+               if (kpage) {
+                       remove_rmap_item_from_tree(tree_rmap_item);
+
+                       lock_page(kpage);
+                       stable_node = stable_tree_insert(kpage);
+                       if (stable_node) {
+                               stable_tree_append(tree_rmap_item, stable_node);
+                               stable_tree_append(rmap_item, stable_node);
+                       }
+                       unlock_page(kpage);
 
                        /*
                         * If we fail to insert the page into the stable tree,
@@ -1140,37 +1247,28 @@ static void cmp_and_merge_page(struct page *page, struct rmap_item *rmap_item)
                         * to a ksm page left outside the stable tree,
                         * in which case we need to break_cow on both.
                         */
-                       if (stable_tree_insert(page2[0], tree_rmap_item))
-                               stable_tree_append(rmap_item, tree_rmap_item);
-                       else {
-                               break_cow(tree_rmap_item->mm,
-                                               tree_rmap_item->address);
-                               break_cow(rmap_item->mm, rmap_item->address);
+                       if (!stable_node) {
+                               break_cow(tree_rmap_item);
+                               break_cow(rmap_item);
                        }
                }
-
-               put_page(page2[0]);
        }
 }
 
 static struct rmap_item *get_next_rmap_item(struct mm_slot *mm_slot,
-                                           struct list_head *cur,
+                                           struct rmap_item **rmap_list,
                                            unsigned long addr)
 {
        struct rmap_item *rmap_item;
 
-       while (cur != &mm_slot->rmap_list) {
-               rmap_item = list_entry(cur, struct rmap_item, link);
-               if ((rmap_item->address & PAGE_MASK) == addr) {
-                       if (!in_stable_tree(rmap_item))
-                               remove_rmap_item_from_tree(rmap_item);
+       while (*rmap_list) {
+               rmap_item = *rmap_list;
+               if ((rmap_item->address & PAGE_MASK) == addr)
                        return rmap_item;
-               }
                if (rmap_item->address > addr)
                        break;
-               cur = cur->next;
+               *rmap_list = rmap_item->rmap_list;
                remove_rmap_item_from_tree(rmap_item);
-               list_del(&rmap_item->link);
                free_rmap_item(rmap_item);
        }
 
@@ -1179,7 +1277,8 @@ static struct rmap_item *get_next_rmap_item(struct mm_slot *mm_slot,
                /* It has already been zeroed */
                rmap_item->mm = mm_slot->mm;
                rmap_item->address = addr;
-               list_add_tail(&rmap_item->link, cur);
+               rmap_item->rmap_list = *rmap_list;
+               *rmap_list = rmap_item;
        }
        return rmap_item;
 }
@@ -1196,21 +1295,43 @@ static struct rmap_item *scan_get_next_rmap_item(struct page **page)
 
        slot = ksm_scan.mm_slot;
        if (slot == &ksm_mm_head) {
+               /*
+                * A number of pages can hang around indefinitely on per-cpu
+                * pagevecs, raised page count preventing write_protect_page
+                * from merging them.  Though it doesn't really matter much,
+                * it is puzzling to see some stuck in pages_volatile until
+                * other activity jostles them out, and they also prevented
+                * LTP's KSM test from succeeding deterministically; so drain
+                * them here (here rather than on entry to ksm_do_scan(),
+                * so we don't IPI too often when pages_to_scan is set low).
+                */
+               lru_add_drain_all();
+
                root_unstable_tree = RB_ROOT;
 
                spin_lock(&ksm_mmlist_lock);
                slot = list_entry(slot->mm_list.next, struct mm_slot, mm_list);
                ksm_scan.mm_slot = slot;
                spin_unlock(&ksm_mmlist_lock);
+               /*
+                * Although we tested list_empty() above, a racing __ksm_exit
+                * of the last mm on the list may have removed it since then.
+                */
+               if (slot == &ksm_mm_head)
+                       return NULL;
 next_mm:
                ksm_scan.address = 0;
-               ksm_scan.rmap_item = list_entry(&slot->rmap_list,
-                                               struct rmap_item, link);
+               ksm_scan.rmap_list = &slot->rmap_list;
        }
 
        mm = slot->mm;
        down_read(&mm->mmap_sem);
-       for (vma = find_vma(mm, ksm_scan.address); vma; vma = vma->vm_next) {
+       if (ksm_test_exit(mm))
+               vma = NULL;
+       else
+               vma = find_vma(mm, ksm_scan.address);
+
+       for (; vma; vma = vma->vm_next) {
                if (!(vma->vm_flags & VM_MERGEABLE))
                        continue;
                if (ksm_scan.address < vma->vm_start)
@@ -1219,62 +1340,76 @@ next_mm:
                        ksm_scan.address = vma->vm_end;
 
                while (ksm_scan.address < vma->vm_end) {
+                       if (ksm_test_exit(mm))
+                               break;
                        *page = follow_page(vma, ksm_scan.address, FOLL_GET);
-                       if (*page && PageAnon(*page)) {
+                       if (IS_ERR_OR_NULL(*page)) {
+                               ksm_scan.address += PAGE_SIZE;
+                               cond_resched();
+                               continue;
+                       }
+                       if (PageAnon(*page) ||
+                           page_trans_compound_anon(*page)) {
                                flush_anon_page(vma, *page, ksm_scan.address);
                                flush_dcache_page(*page);
                                rmap_item = get_next_rmap_item(slot,
-                                       ksm_scan.rmap_item->link.next,
-                                       ksm_scan.address);
+                                       ksm_scan.rmap_list, ksm_scan.address);
                                if (rmap_item) {
-                                       ksm_scan.rmap_item = rmap_item;
+                                       ksm_scan.rmap_list =
+                                                       &rmap_item->rmap_list;
                                        ksm_scan.address += PAGE_SIZE;
                                } else
                                        put_page(*page);
                                up_read(&mm->mmap_sem);
                                return rmap_item;
                        }
-                       if (*page)
-                               put_page(*page);
+                       put_page(*page);
                        ksm_scan.address += PAGE_SIZE;
                        cond_resched();
                }
        }
 
-       if (!ksm_scan.address) {
-               /*
-                * We've completed a full scan of all vmas, holding mmap_sem
-                * throughout, and found no VM_MERGEABLE: so do the same as
-                * __ksm_exit does to remove this mm from all our lists now.
-                */
-               remove_mm_from_lists(mm);
-               up_read(&mm->mmap_sem);
-               slot = ksm_scan.mm_slot;
-               if (slot != &ksm_mm_head)
-                       goto next_mm;
-               return NULL;
+       if (ksm_test_exit(mm)) {
+               ksm_scan.address = 0;
+               ksm_scan.rmap_list = &slot->rmap_list;
        }
-
        /*
         * Nuke all the rmap_items that are above this current rmap:
         * because there were no VM_MERGEABLE vmas with such addresses.
         */
-       remove_trailing_rmap_items(slot, ksm_scan.rmap_item->link.next);
-       up_read(&mm->mmap_sem);
+       remove_trailing_rmap_items(slot, ksm_scan.rmap_list);
 
        spin_lock(&ksm_mmlist_lock);
-       slot = list_entry(slot->mm_list.next, struct mm_slot, mm_list);
-       ksm_scan.mm_slot = slot;
-       spin_unlock(&ksm_mmlist_lock);
+       ksm_scan.mm_slot = list_entry(slot->mm_list.next,
+                                               struct mm_slot, mm_list);
+       if (ksm_scan.address == 0) {
+               /*
+                * We've completed a full scan of all vmas, holding mmap_sem
+                * throughout, and found no VM_MERGEABLE: so do the same as
+                * __ksm_exit does to remove this mm from all our lists now.
+                * This applies either when cleaning up after __ksm_exit
+                * (but beware: we can reach here even before __ksm_exit),
+                * or when all VM_MERGEABLE areas have been unmapped (and
+                * mmap_sem then protects against race with MADV_MERGEABLE).
+                */
+               hlist_del(&slot->link);
+               list_del(&slot->mm_list);
+               spin_unlock(&ksm_mmlist_lock);
+
+               free_mm_slot(slot);
+               clear_bit(MMF_VM_MERGEABLE, &mm->flags);
+               up_read(&mm->mmap_sem);
+               mmdrop(mm);
+       } else {
+               spin_unlock(&ksm_mmlist_lock);
+               up_read(&mm->mmap_sem);
+       }
 
        /* Repeat until we've completed scanning the whole list */
+       slot = ksm_scan.mm_slot;
        if (slot != &ksm_mm_head)
                goto next_mm;
 
-       /*
-        * Bump seqnr here rather than at top, so that __ksm_exit
-        * can skip rb_erase on unstable tree until we run again.
-        */
        ksm_scan.seqnr++;
        return NULL;
 }
@@ -1286,23 +1421,15 @@ next_mm:
 static void ksm_do_scan(unsigned int scan_npages)
 {
        struct rmap_item *rmap_item;
-       struct page *page;
+       struct page *uninitialized_var(page);
 
-       while (scan_npages--) {
+       while (scan_npages-- && likely(!freezing(current))) {
                cond_resched();
                rmap_item = scan_get_next_rmap_item(&page);
                if (!rmap_item)
                        return;
                if (!PageKsm(page) || !in_stable_tree(rmap_item))
                        cmp_and_merge_page(page, rmap_item);
-               else if (page_mapcount(page) == 1) {
-                       /*
-                        * Replace now-unshared ksm page by ordinary page.
-                        */
-                       break_cow(rmap_item->mm, rmap_item->address);
-                       remove_rmap_item_from_tree(rmap_item);
-                       rmap_item->oldchecksum = calc_checksum(page);
-               }
                put_page(page);
        }
 }
@@ -1314,6 +1441,7 @@ static int ksmd_should_run(void)
 
 static int ksm_scan_thread(void *nothing)
 {
+       set_freezable();
        set_user_nice(current, 5);
 
        while (!kthread_should_stop()) {
@@ -1322,11 +1450,13 @@ static int ksm_scan_thread(void *nothing)
                        ksm_do_scan(ksm_thread_pages_to_scan);
                mutex_unlock(&ksm_thread_mutex);
 
+               try_to_freeze();
+
                if (ksmd_should_run()) {
                        schedule_timeout_interruptible(
                                msecs_to_jiffies(ksm_thread_sleep_millisecs));
                } else {
-                       wait_event_interruptible(ksm_thread_wait,
+                       wait_event_freezable(ksm_thread_wait,
                                ksmd_should_run() || kthread_should_stop());
                }
        }
@@ -1346,10 +1476,14 @@ int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
                 */
                if (*vm_flags & (VM_MERGEABLE | VM_SHARED  | VM_MAYSHARE   |
                                 VM_PFNMAP    | VM_IO      | VM_DONTEXPAND |
-                                VM_RESERVED  | VM_HUGETLB | VM_INSERTPAGE |
-                                VM_MIXEDMAP  | VM_SAO))
+                                VM_HUGETLB | VM_NONLINEAR | VM_MIXEDMAP))
                        return 0;               /* just ignore the advice */
 
+#ifdef VM_SAO
+               if (*vm_flags & VM_SAO)
+                       return 0;
+#endif
+
                if (!test_bit(MMF_VM_MERGEABLE, &mm->flags)) {
                        err = __ksm_enter(mm);
                        if (err)
@@ -1399,6 +1533,7 @@ int __ksm_enter(struct mm_struct *mm)
        spin_unlock(&ksm_mmlist_lock);
 
        set_bit(MMF_VM_MERGEABLE, &mm->flags);
+       atomic_inc(&mm->mm_count);
 
        if (needs_wakeup)
                wake_up_interruptible(&ksm_thread_wait);
@@ -1408,16 +1543,301 @@ int __ksm_enter(struct mm_struct *mm)
 
 void __ksm_exit(struct mm_struct *mm)
 {
+       struct mm_slot *mm_slot;
+       int easy_to_free = 0;
+
        /*
-        * This process is exiting: doesn't hold and doesn't need mmap_sem;
-        * but we do need to exclude ksmd and other exiters while we modify
-        * the various lists and trees.
+        * This process is exiting: if it's straightforward (as is the
+        * case when ksmd was never running), free mm_slot immediately.
+        * But if it's at the cursor or has rmap_items linked to it, use
+        * mmap_sem to synchronize with any break_cows before pagetables
+        * are freed, and leave the mm_slot on the list for ksmd to free.
+        * Beware: ksm may already have noticed it exiting and freed the slot.
         */
-       mutex_lock(&ksm_thread_mutex);
-       remove_mm_from_lists(mm);
-       mutex_unlock(&ksm_thread_mutex);
+
+       spin_lock(&ksm_mmlist_lock);
+       mm_slot = get_mm_slot(mm);
+       if (mm_slot && ksm_scan.mm_slot != mm_slot) {
+               if (!mm_slot->rmap_list) {
+                       hlist_del(&mm_slot->link);
+                       list_del(&mm_slot->mm_list);
+                       easy_to_free = 1;
+               } else {
+                       list_move(&mm_slot->mm_list,
+                                 &ksm_scan.mm_slot->mm_list);
+               }
+       }
+       spin_unlock(&ksm_mmlist_lock);
+
+       if (easy_to_free) {
+               free_mm_slot(mm_slot);
+               clear_bit(MMF_VM_MERGEABLE, &mm->flags);
+               mmdrop(mm);
+       } else if (mm_slot) {
+               down_write(&mm->mmap_sem);
+               up_write(&mm->mmap_sem);
+       }
+}
+
+struct page *ksm_does_need_to_copy(struct page *page,
+                       struct vm_area_struct *vma, unsigned long address)
+{
+       struct page *new_page;
+
+       new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, address);
+       if (new_page) {
+               copy_user_highpage(new_page, page, address, vma);
+
+               SetPageDirty(new_page);
+               __SetPageUptodate(new_page);
+               SetPageSwapBacked(new_page);
+               __set_page_locked(new_page);
+
+               if (!mlocked_vma_newpage(vma, new_page))
+                       lru_cache_add_lru(new_page, LRU_ACTIVE_ANON);
+               else
+                       add_page_to_unevictable_list(new_page);
+       }
+
+       return new_page;
 }
 
+int page_referenced_ksm(struct page *page, struct mem_cgroup *memcg,
+                       unsigned long *vm_flags)
+{
+       struct stable_node *stable_node;
+       struct rmap_item *rmap_item;
+       struct hlist_node *hlist;
+       unsigned int mapcount = page_mapcount(page);
+       int referenced = 0;
+       int search_new_forks = 0;
+
+       VM_BUG_ON(!PageKsm(page));
+       VM_BUG_ON(!PageLocked(page));
+
+       stable_node = page_stable_node(page);
+       if (!stable_node)
+               return 0;
+again:
+       hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) {
+               struct anon_vma *anon_vma = rmap_item->anon_vma;
+               struct anon_vma_chain *vmac;
+               struct vm_area_struct *vma;
+
+               anon_vma_lock_read(anon_vma);
+               anon_vma_interval_tree_foreach(vmac, &anon_vma->rb_root,
+                                              0, ULONG_MAX) {
+                       vma = vmac->vma;
+                       if (rmap_item->address < vma->vm_start ||
+                           rmap_item->address >= vma->vm_end)
+                               continue;
+                       /*
+                        * Initially we examine only the vma which covers this
+                        * rmap_item; but later, if there is still work to do,
+                        * we examine covering vmas in other mms: in case they
+                        * were forked from the original since ksmd passed.
+                        */
+                       if ((rmap_item->mm == vma->vm_mm) == search_new_forks)
+                               continue;
+
+                       if (memcg && !mm_match_cgroup(vma->vm_mm, memcg))
+                               continue;
+
+                       referenced += page_referenced_one(page, vma,
+                               rmap_item->address, &mapcount, vm_flags);
+                       if (!search_new_forks || !mapcount)
+                               break;
+               }
+               anon_vma_unlock_read(anon_vma);
+               if (!mapcount)
+                       goto out;
+       }
+       if (!search_new_forks++)
+               goto again;
+out:
+       return referenced;
+}
+
+int try_to_unmap_ksm(struct page *page, enum ttu_flags flags)
+{
+       struct stable_node *stable_node;
+       struct hlist_node *hlist;
+       struct rmap_item *rmap_item;
+       int ret = SWAP_AGAIN;
+       int search_new_forks = 0;
+
+       VM_BUG_ON(!PageKsm(page));
+       VM_BUG_ON(!PageLocked(page));
+
+       stable_node = page_stable_node(page);
+       if (!stable_node)
+               return SWAP_FAIL;
+again:
+       hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) {
+               struct anon_vma *anon_vma = rmap_item->anon_vma;
+               struct anon_vma_chain *vmac;
+               struct vm_area_struct *vma;
+
+               anon_vma_lock_read(anon_vma);
+               anon_vma_interval_tree_foreach(vmac, &anon_vma->rb_root,
+                                              0, ULONG_MAX) {
+                       vma = vmac->vma;
+                       if (rmap_item->address < vma->vm_start ||
+                           rmap_item->address >= vma->vm_end)
+                               continue;
+                       /*
+                        * Initially we examine only the vma which covers this
+                        * rmap_item; but later, if there is still work to do,
+                        * we examine covering vmas in other mms: in case they
+                        * were forked from the original since ksmd passed.
+                        */
+                       if ((rmap_item->mm == vma->vm_mm) == search_new_forks)
+                               continue;
+
+                       ret = try_to_unmap_one(page, vma,
+                                       rmap_item->address, flags);
+                       if (ret != SWAP_AGAIN || !page_mapped(page)) {
+                               anon_vma_unlock_read(anon_vma);
+                               goto out;
+                       }
+               }
+               anon_vma_unlock_read(anon_vma);
+       }
+       if (!search_new_forks++)
+               goto again;
+out:
+       return ret;
+}
+
+#ifdef CONFIG_MIGRATION
+int rmap_walk_ksm(struct page *page, int (*rmap_one)(struct page *,
+                 struct vm_area_struct *, unsigned long, void *), void *arg)
+{
+       struct stable_node *stable_node;
+       struct hlist_node *hlist;
+       struct rmap_item *rmap_item;
+       int ret = SWAP_AGAIN;
+       int search_new_forks = 0;
+
+       VM_BUG_ON(!PageKsm(page));
+       VM_BUG_ON(!PageLocked(page));
+
+       stable_node = page_stable_node(page);
+       if (!stable_node)
+               return ret;
+again:
+       hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) {
+               struct anon_vma *anon_vma = rmap_item->anon_vma;
+               struct anon_vma_chain *vmac;
+               struct vm_area_struct *vma;
+
+               anon_vma_lock_read(anon_vma);
+               anon_vma_interval_tree_foreach(vmac, &anon_vma->rb_root,
+                                              0, ULONG_MAX) {
+                       vma = vmac->vma;
+                       if (rmap_item->address < vma->vm_start ||
+                           rmap_item->address >= vma->vm_end)
+                               continue;
+                       /*
+                        * Initially we examine only the vma which covers this
+                        * rmap_item; but later, if there is still work to do,
+                        * we examine covering vmas in other mms: in case they
+                        * were forked from the original since ksmd passed.
+                        */
+                       if ((rmap_item->mm == vma->vm_mm) == search_new_forks)
+                               continue;
+
+                       ret = rmap_one(page, vma, rmap_item->address, arg);
+                       if (ret != SWAP_AGAIN) {
+                               anon_vma_unlock_read(anon_vma);
+                               goto out;
+                       }
+               }
+               anon_vma_unlock_read(anon_vma);
+       }
+       if (!search_new_forks++)
+               goto again;
+out:
+       return ret;
+}
+
+void ksm_migrate_page(struct page *newpage, struct page *oldpage)
+{
+       struct stable_node *stable_node;
+
+       VM_BUG_ON(!PageLocked(oldpage));
+       VM_BUG_ON(!PageLocked(newpage));
+       VM_BUG_ON(newpage->mapping != oldpage->mapping);
+
+       stable_node = page_stable_node(newpage);
+       if (stable_node) {
+               VM_BUG_ON(stable_node->kpfn != page_to_pfn(oldpage));
+               stable_node->kpfn = page_to_pfn(newpage);
+       }
+}
+#endif /* CONFIG_MIGRATION */
+
+#ifdef CONFIG_MEMORY_HOTREMOVE
+static struct stable_node *ksm_check_stable_tree(unsigned long start_pfn,
+                                                unsigned long end_pfn)
+{
+       struct rb_node *node;
+
+       for (node = rb_first(&root_stable_tree); node; node = rb_next(node)) {
+               struct stable_node *stable_node;
+
+               stable_node = rb_entry(node, struct stable_node, node);
+               if (stable_node->kpfn >= start_pfn &&
+                   stable_node->kpfn < end_pfn)
+                       return stable_node;
+       }
+       return NULL;
+}
+
+static int ksm_memory_callback(struct notifier_block *self,
+                              unsigned long action, void *arg)
+{
+       struct memory_notify *mn = arg;
+       struct stable_node *stable_node;
+
+       switch (action) {
+       case MEM_GOING_OFFLINE:
+               /*
+                * Keep it very simple for now: just lock out ksmd and
+                * MADV_UNMERGEABLE while any memory is going offline.
+                * mutex_lock_nested() is necessary because lockdep was alarmed
+                * that here we take ksm_thread_mutex inside notifier chain
+                * mutex, and later take notifier chain mutex inside
+                * ksm_thread_mutex to unlock it.   But that's safe because both
+                * are inside mem_hotplug_mutex.
+                */
+               mutex_lock_nested(&ksm_thread_mutex, SINGLE_DEPTH_NESTING);
+               break;
+
+       case MEM_OFFLINE:
+               /*
+                * Most of the work is done by page migration; but there might
+                * be a few stable_nodes left over, still pointing to struct
+                * pages which have been offlined: prune those from the tree.
+                */
+               while ((stable_node = ksm_check_stable_tree(mn->start_pfn,
+                                       mn->start_pfn + mn->nr_pages)) != NULL)
+                       remove_node_from_stable_tree(stable_node);
+               /* fallthrough */
+
+       case MEM_CANCEL_OFFLINE:
+               mutex_unlock(&ksm_thread_mutex);
+               break;
+       }
+       return NOTIFY_OK;
+}
+#endif /* CONFIG_MEMORY_HOTREMOVE */
+
+#ifdef CONFIG_SYSFS
+/*
+ * This all compiles without CONFIG_SYSFS, but is a waste of space.
+ */
+
 #define KSM_ATTR_RO(_name) \
        static struct kobj_attribute _name##_attr = __ATTR_RO(_name)
 #define KSM_ATTR(_name) \
@@ -1491,15 +1911,17 @@ static ssize_t run_store(struct kobject *kobj, struct kobj_attribute *attr,
        /*
         * KSM_RUN_MERGE sets ksmd running, and 0 stops it running.
         * KSM_RUN_UNMERGE stops it running and unmerges all rmap_items,
-        * breaking COW to free the unswappable pages_shared (but leaves
-        * mm_slots on the list for when ksmd may be set running again).
+        * breaking COW to free the pages_shared (but leaves mm_slots
+        * on the list for when ksmd may be set running again).
         */
 
        mutex_lock(&ksm_thread_mutex);
        if (ksm_run != flags) {
                ksm_run = flags;
                if (flags & KSM_RUN_UNMERGE) {
+                       set_current_oom_origin();
                        err = unmerge_and_remove_all_rmap_items();
+                       clear_current_oom_origin();
                        if (err) {
                                ksm_run = KSM_RUN_STOP;
                                count = err;
@@ -1515,29 +1937,6 @@ static ssize_t run_store(struct kobject *kobj, struct kobj_attribute *attr,
 }
 KSM_ATTR(run);
 
-static ssize_t max_kernel_pages_store(struct kobject *kobj,
-                                     struct kobj_attribute *attr,
-                                     const char *buf, size_t count)
-{
-       int err;
-       unsigned long nr_pages;
-
-       err = strict_strtoul(buf, 10, &nr_pages);
-       if (err)
-               return -EINVAL;
-
-       ksm_max_kernel_pages = nr_pages;
-
-       return count;
-}
-
-static ssize_t max_kernel_pages_show(struct kobject *kobj,
-                                    struct kobj_attribute *attr, char *buf)
-{
-       return sprintf(buf, "%lu\n", ksm_max_kernel_pages);
-}
-KSM_ATTR(max_kernel_pages);
-
 static ssize_t pages_shared_show(struct kobject *kobj,
                                 struct kobj_attribute *attr, char *buf)
 {
@@ -1587,7 +1986,6 @@ static struct attribute *ksm_attrs[] = {
        &sleep_millisecs_attr.attr,
        &pages_to_scan_attr.attr,
        &run_attr.attr,
-       &max_kernel_pages_attr.attr,
        &pages_shared_attr.attr,
        &pages_sharing_attr.attr,
        &pages_unshared_attr.attr,
@@ -1600,6 +1998,7 @@ static struct attribute_group ksm_attr_group = {
        .attrs = ksm_attrs,
        .name = "ksm",
 };
+#endif /* CONFIG_SYSFS */
 
 static int __init ksm_init(void)
 {
@@ -1610,30 +2009,35 @@ static int __init ksm_init(void)
        if (err)
                goto out;
 
-       err = mm_slots_hash_init();
-       if (err)
-               goto out_free1;
-
        ksm_thread = kthread_run(ksm_scan_thread, NULL, "ksmd");
        if (IS_ERR(ksm_thread)) {
                printk(KERN_ERR "ksm: creating kthread failed\n");
                err = PTR_ERR(ksm_thread);
-               goto out_free2;
+               goto out_free;
        }
 
+#ifdef CONFIG_SYSFS
        err = sysfs_create_group(mm_kobj, &ksm_attr_group);
        if (err) {
                printk(KERN_ERR "ksm: register sysfs failed\n");
-               goto out_free3;
+               kthread_stop(ksm_thread);
+               goto out_free;
        }
+#else
+       ksm_run = KSM_RUN_MERGE;        /* no way for user to start it */
 
+#endif /* CONFIG_SYSFS */
+
+#ifdef CONFIG_MEMORY_HOTREMOVE
+       /*
+        * Choose a high priority since the callback takes ksm_thread_mutex:
+        * later callbacks could only be taking locks which nest within that.
+        */
+       hotplug_memory_notifier(ksm_memory_callback, 100);
+#endif
        return 0;
 
-out_free3:
-       kthread_stop(ksm_thread);
-out_free2:
-       mm_slots_hash_free();
-out_free1:
+out_free:
        ksm_slab_free();
 out:
        return err;