1 files changed, 76 insertions, 37 deletions

diff --git a/mm/swapfile.c b/mm/swapfile.c
index f071648..eade24d 100644
--- a/mm/swapfile.c
+++ b/mm/swapfile.c
@@ -506,9 +506,19 @@ unsigned int count_swap_pages(int type, int free)
  * just let do_wp_page work it out if a write is requested later - to
  * force COW, vm_page_prot omits write permission from any private vma.
  */
-static void unuse_pte(struct vm_area_struct *vma, pte_t *pte,
+static int unuse_pte(struct vm_area_struct *vma, pmd_t *pmd,
 		unsigned long addr, swp_entry_t entry, struct page *page)
 {
+	spinlock_t *ptl;
+	pte_t *pte;
+	int found = 1;
+
+	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
+	if (unlikely(!pte_same(*pte, swp_entry_to_pte(entry)))) {
+		found = 0;
+		goto out;
+	}
+
 	inc_mm_counter(vma->vm_mm, anon_rss);
 	get_page(page);
 	set_pte_at(vma->vm_mm, addr, pte,
@@ -520,6 +530,9 @@ static void unuse_pte(struct vm_area_struct *vma, pte_t *pte,
 	 * immediately swapped out again after swapon.
 	 */
 	activate_page(page);
+out:
+	pte_unmap_unlock(pte, ptl);
+	return found;
 }
 
 static int unuse_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
@@ -528,22 +541,33 @@ static int unuse_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
 {
 	pte_t swp_pte = swp_entry_to_pte(entry);
 	pte_t *pte;
-	spinlock_t *ptl;
 	int found = 0;
 
-	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
+	/*
+	 * We don't actually need pte lock while scanning for swp_pte: since
+	 * we hold page lock and mmap_sem, swp_pte cannot be inserted into the
+	 * page table while we're scanning; though it could get zapped, and on
+	 * some architectures (e.g. x86_32 with PAE) we might catch a glimpse
+	 * of unmatched parts which look like swp_pte, so unuse_pte must
+	 * recheck under pte lock.  Scanning without pte lock lets it be
+	 * preemptible whenever CONFIG_PREEMPT but not CONFIG_HIGHPTE.
+	 */
+	pte = pte_offset_map(pmd, addr);
 	do {
 		/*
 		 * swapoff spends a _lot_ of time in this loop!
 		 * Test inline before going to call unuse_pte.
 		 */
 		if (unlikely(pte_same(*pte, swp_pte))) {
-			unuse_pte(vma, pte++, addr, entry, page);
-			found = 1;
-			break;
+			pte_unmap(pte);
+			found = unuse_pte(vma, pmd, addr, entry, page);
+			if (found)
+				goto out;
+			pte = pte_offset_map(pmd, addr);
 		}
 	} while (pte++, addr += PAGE_SIZE, addr != end);
-	pte_unmap_unlock(pte - 1, ptl);
+	pte_unmap(pte - 1);
+out:
 	return found;
 }
 
@@ -730,7 +754,8 @@ static int try_to_unuse(unsigned int type)
 		 */
 		swap_map = &si->swap_map[i];
 		entry = swp_entry(type, i);
-		page = read_swap_cache_async(entry, NULL, 0);
+		page = read_swap_cache_async(entry,
+					GFP_HIGHUSER_MOVABLE, NULL, 0);
 		if (!page) {
 			/*
 			 * Either swap_duplicate() failed because entry
@@ -789,7 +814,7 @@ static int try_to_unuse(unsigned int type)
 			atomic_inc(&new_start_mm->mm_users);
 			atomic_inc(&prev_mm->mm_users);
 			spin_lock(&mmlist_lock);
-			while (*swap_map > 1 && !retval &&
+			while (*swap_map > 1 && !retval && !shmem &&
 					(p = p->next) != &start_mm->mmlist) {
 				mm = list_entry(p, struct mm_struct, mmlist);
 				if (!atomic_inc_not_zero(&mm->mm_users))
@@ -821,6 +846,13 @@ static int try_to_unuse(unsigned int type)
 			mmput(start_mm);
 			start_mm = new_start_mm;
 		}
+		if (shmem) {
+			/* page has already been unlocked and released */
+			if (shmem > 0)
+				continue;
+			retval = shmem;
+			break;
+		}
 		if (retval) {
 			unlock_page(page);
 			page_cache_release(page);
@@ -859,12 +891,6 @@ static int try_to_unuse(unsigned int type)
 		 * read from disk into another page.  Splitting into two
 		 * pages would be incorrect if swap supported "shared
 		 * private" pages, but they are handled by tmpfs files.
-		 *
-		 * Note shmem_unuse already deleted a swappage from
-		 * the swap cache, unless the move to filepage failed:
-		 * in which case it left swappage in cache, lowered its
-		 * swap count to pass quickly through the loops above,
-		 * and now we must reincrement count to try again later.
 		 */
 		if ((*swap_map > 1) && PageDirty(page) && PageSwapCache(page)) {
 			struct writeback_control wbc = {
@@ -875,12 +901,8 @@ static int try_to_unuse(unsigned int type)
 			lock_page(page);
 			wait_on_page_writeback(page);
 		}
-		if (PageSwapCache(page)) {
-			if (shmem)
-				swap_duplicate(entry);
-			else
-				delete_from_swap_cache(page);
-		}
+		if (PageSwapCache(page))
+			delete_from_swap_cache(page);
 
 		/*
 		 * So we could skip searching mms once swap count went
@@ -1768,31 +1790,48 @@ get_swap_info_struct(unsigned type)
  */
 int valid_swaphandles(swp_entry_t entry, unsigned long *offset)
 {
+	struct swap_info_struct *si;
 	int our_page_cluster = page_cluster;
-	int ret = 0, i = 1 << our_page_cluster;
-	unsigned long toff;
-	struct swap_info_struct *swapdev = swp_type(entry) + swap_info;
+	pgoff_t target, toff;
+	pgoff_t base, end;
+	int nr_pages = 0;
 
 	if (!our_page_cluster)	/* no readahead */
 		return 0;
-	toff = (swp_offset(entry) >> our_page_cluster) << our_page_cluster;
-	if (!toff)		/* first page is swap header */
-		toff++, i--;
-	*offset = toff;
+
+	si = &swap_info[swp_type(entry)];
+	target = swp_offset(entry);
+	base = (target >> our_page_cluster) << our_page_cluster;
+	end = base + (1 << our_page_cluster);
+	if (!base)		/* first page is swap header */
+		base++;
 
 	spin_lock(&swap_lock);
-	do {
-		/* Don't read-ahead past the end of the swap area */
-		if (toff >= swapdev->max)
+	if (end > si->max)	/* don't go beyond end of map */
+		end = si->max;
+
+	/* Count contiguous allocated slots above our target */
+	for (toff = target; ++toff < end; nr_pages++) {
+		/* Don't read in free or bad pages */
+		if (!si->swap_map[toff])
+			break;
+		if (si->swap_map[toff] == SWAP_MAP_BAD)
 			break;
+	}
+	/* Count contiguous allocated slots below our target */
+	for (toff = target; --toff >= base; nr_pages++) {
 		/* Don't read in free or bad pages */
-		if (!swapdev->swap_map[toff])
+		if (!si->swap_map[toff])
 			break;
-		if (swapdev->swap_map[toff] == SWAP_MAP_BAD)
+		if (si->swap_map[toff] == SWAP_MAP_BAD)
 			break;
-		toff++;
-		ret++;
-	} while (--i);
+	}
 	spin_unlock(&swap_lock);
-	return ret;
+
+	/*
+	 * Indicate starting offset, and return number of pages to get:
+	 * if only 1, say 0, since there's then no readahead to be done.
+	 */
+	*offset = ++toff;
+	return nr_pages? ++nr_pages: 0;
 }