[PATCH] smaps: shared fix
authorNick Piggin <nickpiggin@yahoo.com.au>
Mon, 6 Mar 2006 23:42:58 +0000 (15:42 -0800)
committerLinus Torvalds <torvalds@g5.osdl.org>
Tue, 7 Mar 2006 02:40:45 +0000 (18:40 -0800)
The point of the smaps "shared" is to count the number of pages that are
mapped by more than one process, according to Mauricio Lin.  However, smaps
uses page_count for this, so it will return a false positive for every page
that is mapped by just that one process, which is also in pagecache or
swapcache.  There are false positive situations for anonymous pages not in
swapcache as well: - page reclaim, migration - get_user_pages (eg.
direct-io, ptrace)

Use page_mapcount instead, to count the number of mappings to the page.

Use vm_normal_page so that weird things like /dev/mem aren't counted either.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
fs/proc/task_mmu.c

index 35787f907f1290c6982759e084d87c5b601cc6c5..91b7c15ab3730c4f17ffc4a2f4bdb5e22da54edf 100644 (file)
@@ -204,7 +204,6 @@ static void smaps_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
 {
        pte_t *pte, ptent;
        spinlock_t *ptl;
-       unsigned long pfn;
        struct page *page;
 
        pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
@@ -214,12 +213,12 @@ static void smaps_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
                        continue;
 
                mss->resident += PAGE_SIZE;
-               pfn = pte_pfn(ptent);
-               if (!pfn_valid(pfn))
+
+               page = vm_normal_page(vma, addr, ptent);
+               if (!page)
                        continue;
 
-               page = pfn_to_page(pfn);
-               if (page_count(page) >= 2) {
+               if (page_mapcount(page) >= 2) {
                        if (pte_dirty(ptent))
                                mss->shared_dirty += PAGE_SIZE;
                        else