mincore: pass ranges as start,end address pairs
authorJohannes Weiner <hannes@cmpxchg.org>
Mon, 24 May 2010 21:32:11 +0000 (14:32 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 25 May 2010 15:06:58 +0000 (08:06 -0700)
Instead of passing a start address and a number of pages into the helper
functions, convert them to use a start and an end address.

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/mincore.c

index a0c4c10bbab7bec64ea525d6e57f51da5aa879c1..211604adc23ccfb52ef4313534f5ad0fde2173db 100644 (file)
 #include <asm/pgtable.h>
 
 static void mincore_hugetlb_page_range(struct vm_area_struct *vma,
-                               unsigned long addr, unsigned long nr,
+                               unsigned long addr, unsigned long end,
                                unsigned char *vec)
 {
 #ifdef CONFIG_HUGETLB_PAGE
        struct hstate *h;
-       int i;
 
-       i = 0;
        h = hstate_vma(vma);
        while (1) {
                unsigned char present;
@@ -40,10 +38,10 @@ static void mincore_hugetlb_page_range(struct vm_area_struct *vma,
                                       addr & huge_page_mask(h));
                present = ptep && !huge_pte_none(huge_ptep_get(ptep));
                while (1) {
-                       vec[i++] = present;
+                       *vec = present;
+                       vec++;
                        addr += PAGE_SIZE;
-                       /* reach buffer limit */
-                       if (i == nr)
+                       if (addr == end)
                                return;
                        /* check hugepage border */
                        if (!(addr & ~huge_page_mask(h)))
@@ -86,9 +84,10 @@ static unsigned char mincore_page(struct address_space *mapping, pgoff_t pgoff)
 }
 
 static void mincore_unmapped_range(struct vm_area_struct *vma,
-                               unsigned long addr, unsigned long nr,
+                               unsigned long addr, unsigned long end,
                                unsigned char *vec)
 {
+       unsigned long nr = (end - addr) >> PAGE_SHIFT;
        int i;
 
        if (vma->vm_file) {
@@ -104,42 +103,44 @@ static void mincore_unmapped_range(struct vm_area_struct *vma,
 }
 
 static void mincore_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
-                       unsigned long addr, unsigned long nr,
+                       unsigned long addr, unsigned long end,
                        unsigned char *vec)
 {
+       unsigned long next;
        spinlock_t *ptl;
        pte_t *ptep;
-       int i;
 
        ptep = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
-       for (i = 0; i < nr; i++, ptep++, addr += PAGE_SIZE) {
+       do {
                pte_t pte = *ptep;
                pgoff_t pgoff;
 
+               next = addr + PAGE_SIZE;
                if (pte_none(pte))
-                       mincore_unmapped_range(vma, addr, 1, vec);
+                       mincore_unmapped_range(vma, addr, next, vec);
                else if (pte_present(pte))
-                       vec[i] = 1;
+                       *vec = 1;
                else if (pte_file(pte)) {
                        pgoff = pte_to_pgoff(pte);
-                       vec[i] = mincore_page(vma->vm_file->f_mapping, pgoff);
+                       *vec = mincore_page(vma->vm_file->f_mapping, pgoff);
                } else { /* pte is a swap entry */
                        swp_entry_t entry = pte_to_swp_entry(pte);
 
                        if (is_migration_entry(entry)) {
                                /* migration entries are always uptodate */
-                               vec[i] = 1;
+                               *vec = 1;
                        } else {
 #ifdef CONFIG_SWAP
                                pgoff = entry.val;
-                               vec[i] = mincore_page(&swapper_space, pgoff);
+                               *vec = mincore_page(&swapper_space, pgoff);
 #else
                                WARN_ON(1);
-                               vec[i] = 1;
+                               *vec = 1;
 #endif
                        }
                }
-       }
+               vec++;
+       } while (ptep++, addr = next, addr != end);
        pte_unmap_unlock(ptep - 1, ptl);
 }
 
@@ -153,25 +154,21 @@ static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *v
        pgd_t *pgd;
        pud_t *pud;
        pmd_t *pmd;
-       unsigned long nr;
        struct vm_area_struct *vma;
+       unsigned long end;
 
        vma = find_vma(current->mm, addr);
        if (!vma || addr < vma->vm_start)
                return -ENOMEM;
 
-       nr = min(pages, (vma->vm_end - addr) >> PAGE_SHIFT);
+       end = min(vma->vm_end, addr + (pages << PAGE_SHIFT));
 
        if (is_vm_hugetlb_page(vma)) {
-               mincore_hugetlb_page_range(vma, addr, nr, vec);
-               return nr;
+               mincore_hugetlb_page_range(vma, addr, end, vec);
+               return (end - addr) >> PAGE_SHIFT;
        }
 
-       /*
-        * Calculate how many pages there are left in the last level of the
-        * PTE array for our address.
-        */
-       nr = min(nr, PTRS_PER_PTE - ((addr >> PAGE_SHIFT) & (PTRS_PER_PTE-1)));
+       end = pmd_addr_end(addr, end);
 
        pgd = pgd_offset(vma->vm_mm, addr);
        if (pgd_none_or_clear_bad(pgd))
@@ -183,12 +180,12 @@ static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *v
        if (pmd_none_or_clear_bad(pmd))
                goto none_mapped;
 
-       mincore_pte_range(vma, pmd, addr, nr, vec);
-       return nr;
+       mincore_pte_range(vma, pmd, addr, end, vec);
+       return (end - addr) >> PAGE_SHIFT;
 
 none_mapped:
-       mincore_unmapped_range(vma, addr, nr, vec);
-       return nr;
+       mincore_unmapped_range(vma, addr, end, vec);
+       return (end - addr) >> PAGE_SHIFT;
 }
 
 /*