mm: shrinker: make shrinker not depend on memcg kmem
[linux-2.6-block.git] / mm / pagewalk.c
index c3084ff2569d2f297ed369874ba839aadc93785e..d48c2a986ea3f8dfffaed7c2c99ca980375f4726 100644 (file)
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: GPL-2.0
-#include <linux/mm.h>
+#include <linux/pagewalk.h>
 #include <linux/highmem.h>
 #include <linux/sched.h>
 #include <linux/hugetlb.h>
@@ -9,10 +9,11 @@ static int walk_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
 {
        pte_t *pte;
        int err = 0;
+       const struct mm_walk_ops *ops = walk->ops;
 
        pte = pte_offset_map(pmd, addr);
        for (;;) {
-               err = walk->pte_entry(pte, addr, addr + PAGE_SIZE, walk);
+               err = ops->pte_entry(pte, addr, addr + PAGE_SIZE, walk);
                if (err)
                       break;
                addr += PAGE_SIZE;
@@ -30,6 +31,7 @@ static int walk_pmd_range(pud_t *pud, unsigned long addr, unsigned long end,
 {
        pmd_t *pmd;
        unsigned long next;
+       const struct mm_walk_ops *ops = walk->ops;
        int err = 0;
 
        pmd = pmd_offset(pud, addr);
@@ -37,8 +39,8 @@ static int walk_pmd_range(pud_t *pud, unsigned long addr, unsigned long end,
 again:
                next = pmd_addr_end(addr, end);
                if (pmd_none(*pmd) || !walk->vma) {
-                       if (walk->pte_hole)
-                               err = walk->pte_hole(addr, next, walk);
+                       if (ops->pte_hole)
+                               err = ops->pte_hole(addr, next, walk);
                        if (err)
                                break;
                        continue;
@@ -47,8 +49,8 @@ again:
                 * This implies that each ->pmd_entry() handler
                 * needs to know about pmd_trans_huge() pmds
                 */
-               if (walk->pmd_entry)
-                       err = walk->pmd_entry(pmd, addr, next, walk);
+               if (ops->pmd_entry)
+                       err = ops->pmd_entry(pmd, addr, next, walk);
                if (err)
                        break;
 
@@ -56,7 +58,7 @@ again:
                 * Check this here so we only break down trans_huge
                 * pages when we _need_ to
                 */
-               if (!walk->pte_entry)
+               if (!ops->pte_entry)
                        continue;
 
                split_huge_pmd(walk->vma, pmd, addr);
@@ -75,6 +77,7 @@ static int walk_pud_range(p4d_t *p4d, unsigned long addr, unsigned long end,
 {
        pud_t *pud;
        unsigned long next;
+       const struct mm_walk_ops *ops = walk->ops;
        int err = 0;
 
        pud = pud_offset(p4d, addr);
@@ -82,18 +85,18 @@ static int walk_pud_range(p4d_t *p4d, unsigned long addr, unsigned long end,
  again:
                next = pud_addr_end(addr, end);
                if (pud_none(*pud) || !walk->vma) {
-                       if (walk->pte_hole)
-                               err = walk->pte_hole(addr, next, walk);
+                       if (ops->pte_hole)
+                               err = ops->pte_hole(addr, next, walk);
                        if (err)
                                break;
                        continue;
                }
 
-               if (walk->pud_entry) {
+               if (ops->pud_entry) {
                        spinlock_t *ptl = pud_trans_huge_lock(pud, walk->vma);
 
                        if (ptl) {
-                               err = walk->pud_entry(pud, addr, next, walk);
+                               err = ops->pud_entry(pud, addr, next, walk);
                                spin_unlock(ptl);
                                if (err)
                                        break;
@@ -105,7 +108,7 @@ static int walk_pud_range(p4d_t *p4d, unsigned long addr, unsigned long end,
                if (pud_none(*pud))
                        goto again;
 
-               if (walk->pmd_entry || walk->pte_entry)
+               if (ops->pmd_entry || ops->pte_entry)
                        err = walk_pmd_range(pud, addr, next, walk);
                if (err)
                        break;
@@ -119,19 +122,20 @@ static int walk_p4d_range(pgd_t *pgd, unsigned long addr, unsigned long end,
 {
        p4d_t *p4d;
        unsigned long next;
+       const struct mm_walk_ops *ops = walk->ops;
        int err = 0;
 
        p4d = p4d_offset(pgd, addr);
        do {
                next = p4d_addr_end(addr, end);
                if (p4d_none_or_clear_bad(p4d)) {
-                       if (walk->pte_hole)
-                               err = walk->pte_hole(addr, next, walk);
+                       if (ops->pte_hole)
+                               err = ops->pte_hole(addr, next, walk);
                        if (err)
                                break;
                        continue;
                }
-               if (walk->pmd_entry || walk->pte_entry)
+               if (ops->pmd_entry || ops->pte_entry)
                        err = walk_pud_range(p4d, addr, next, walk);
                if (err)
                        break;
@@ -145,19 +149,20 @@ static int walk_pgd_range(unsigned long addr, unsigned long end,
 {
        pgd_t *pgd;
        unsigned long next;
+       const struct mm_walk_ops *ops = walk->ops;
        int err = 0;
 
        pgd = pgd_offset(walk->mm, addr);
        do {
                next = pgd_addr_end(addr, end);
                if (pgd_none_or_clear_bad(pgd)) {
-                       if (walk->pte_hole)
-                               err = walk->pte_hole(addr, next, walk);
+                       if (ops->pte_hole)
+                               err = ops->pte_hole(addr, next, walk);
                        if (err)
                                break;
                        continue;
                }
-               if (walk->pmd_entry || walk->pte_entry)
+               if (ops->pmd_entry || ops->pte_entry)
                        err = walk_p4d_range(pgd, addr, next, walk);
                if (err)
                        break;
@@ -183,6 +188,7 @@ static int walk_hugetlb_range(unsigned long addr, unsigned long end,
        unsigned long hmask = huge_page_mask(h);
        unsigned long sz = huge_page_size(h);
        pte_t *pte;
+       const struct mm_walk_ops *ops = walk->ops;
        int err = 0;
 
        do {
@@ -190,9 +196,9 @@ static int walk_hugetlb_range(unsigned long addr, unsigned long end,
                pte = huge_pte_offset(walk->mm, addr & hmask, sz);
 
                if (pte)
-                       err = walk->hugetlb_entry(pte, hmask, addr, next, walk);
-               else if (walk->pte_hole)
-                       err = walk->pte_hole(addr, next, walk);
+                       err = ops->hugetlb_entry(pte, hmask, addr, next, walk);
+               else if (ops->pte_hole)
+                       err = ops->pte_hole(addr, next, walk);
 
                if (err)
                        break;
@@ -220,9 +226,10 @@ static int walk_page_test(unsigned long start, unsigned long end,
                        struct mm_walk *walk)
 {
        struct vm_area_struct *vma = walk->vma;
+       const struct mm_walk_ops *ops = walk->ops;
 
-       if (walk->test_walk)
-               return walk->test_walk(start, end, walk);
+       if (ops->test_walk)
+               return ops->test_walk(start, end, walk);
 
        /*
         * vma(VM_PFNMAP) doesn't have any valid struct pages behind VM_PFNMAP
@@ -234,8 +241,8 @@ static int walk_page_test(unsigned long start, unsigned long end,
         */
        if (vma->vm_flags & VM_PFNMAP) {
                int err = 1;
-               if (walk->pte_hole)
-                       err = walk->pte_hole(start, end, walk);
+               if (ops->pte_hole)
+                       err = ops->pte_hole(start, end, walk);
                return err ? err : 1;
        }
        return 0;
@@ -248,7 +255,7 @@ static int __walk_page_range(unsigned long start, unsigned long end,
        struct vm_area_struct *vma = walk->vma;
 
        if (vma && is_vm_hugetlb_page(vma)) {
-               if (walk->hugetlb_entry)
+               if (walk->ops->hugetlb_entry)
                        err = walk_hugetlb_range(start, end, walk);
        } else
                err = walk_pgd_range(start, end, walk);
@@ -258,11 +265,13 @@ static int __walk_page_range(unsigned long start, unsigned long end,
 
 /**
  * walk_page_range - walk page table with caller specific callbacks
- * @start: start address of the virtual address range
- * @end: end address of the virtual address range
- * @walk: mm_walk structure defining the callbacks and the target address space
+ * @mm:                mm_struct representing the target process of page table walk
+ * @start:     start address of the virtual address range
+ * @end:       end address of the virtual address range
+ * @ops:       operation to call during the walk
+ * @private:   private data for callbacks' usage
  *
- * Recursively walk the page table tree of the process represented by @walk->mm
+ * Recursively walk the page table tree of the process represented by @mm
  * within the virtual address range [@start, @end). During walking, we can do
  * some caller-specific works for each entry, by setting up pmd_entry(),
  * pte_entry(), and/or hugetlb_entry(). If you don't set up for some of these
@@ -278,47 +287,52 @@ static int __walk_page_range(unsigned long start, unsigned long end,
  *
  * Before starting to walk page table, some callers want to check whether
  * they really want to walk over the current vma, typically by checking
- * its vm_flags. walk_page_test() and @walk->test_walk() are used for this
+ * its vm_flags. walk_page_test() and @ops->test_walk() are used for this
  * purpose.
  *
  * struct mm_walk keeps current values of some common data like vma and pmd,
  * which are useful for the access from callbacks. If you want to pass some
- * caller-specific data to callbacks, @walk->private should be helpful.
+ * caller-specific data to callbacks, @private should be helpful.
  *
  * Locking:
- *   Callers of walk_page_range() and walk_page_vma() should hold
- *   @walk->mm->mmap_sem, because these function traverse vma list and/or
- *   access to vma's data.
+ *   Callers of walk_page_range() and walk_page_vma() should hold @mm->mmap_sem,
+ *   because these function traverse vma list and/or access to vma's data.
  */
-int walk_page_range(unsigned long start, unsigned long end,
-                   struct mm_walk *walk)
+int walk_page_range(struct mm_struct *mm, unsigned long start,
+               unsigned long end, const struct mm_walk_ops *ops,
+               void *private)
 {
        int err = 0;
        unsigned long next;
        struct vm_area_struct *vma;
+       struct mm_walk walk = {
+               .ops            = ops,
+               .mm             = mm,
+               .private        = private,
+       };
 
        if (start >= end)
                return -EINVAL;
 
-       if (!walk->mm)
+       if (!walk.mm)
                return -EINVAL;
 
-       VM_BUG_ON_MM(!rwsem_is_locked(&walk->mm->mmap_sem), walk->mm);
+       lockdep_assert_held(&walk.mm->mmap_sem);
 
-       vma = find_vma(walk->mm, start);
+       vma = find_vma(walk.mm, start);
        do {
                if (!vma) { /* after the last vma */
-                       walk->vma = NULL;
+                       walk.vma = NULL;
                        next = end;
                } else if (start < vma->vm_start) { /* outside vma */
-                       walk->vma = NULL;
+                       walk.vma = NULL;
                        next = min(end, vma->vm_start);
                } else { /* inside vma */
-                       walk->vma = vma;
+                       walk.vma = vma;
                        next = min(end, vma->vm_end);
                        vma = vma->vm_next;
 
-                       err = walk_page_test(start, next, walk);
+                       err = walk_page_test(start, next, &walk);
                        if (err > 0) {
                                /*
                                 * positive return values are purely for
@@ -331,28 +345,34 @@ int walk_page_range(unsigned long start, unsigned long end,
                        if (err < 0)
                                break;
                }
-               if (walk->vma || walk->pte_hole)
-                       err = __walk_page_range(start, next, walk);
+               if (walk.vma || walk.ops->pte_hole)
+                       err = __walk_page_range(start, next, &walk);
                if (err)
                        break;
        } while (start = next, start < end);
        return err;
 }
 
-int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk)
+int walk_page_vma(struct vm_area_struct *vma, const struct mm_walk_ops *ops,
+               void *private)
 {
+       struct mm_walk walk = {
+               .ops            = ops,
+               .mm             = vma->vm_mm,
+               .vma            = vma,
+               .private        = private,
+       };
        int err;
 
-       if (!walk->mm)
+       if (!walk.mm)
                return -EINVAL;
 
-       VM_BUG_ON(!rwsem_is_locked(&walk->mm->mmap_sem));
-       VM_BUG_ON(!vma);
-       walk->vma = vma;
-       err = walk_page_test(vma->vm_start, vma->vm_end, walk);
+       lockdep_assert_held(&walk.mm->mmap_sem);
+
+       err = walk_page_test(vma->vm_start, vma->vm_end, &walk);
        if (err > 0)
                return 0;
        if (err < 0)
                return err;
-       return __walk_page_range(vma->vm_start, vma->vm_end, walk);
+       return __walk_page_range(vma->vm_start, vma->vm_end, &walk);
 }