mm: account skipped entries to avoid looping in find_get_pages
Shaohua Li [Thu, 15 Sep 2011 00:45:19 +0000 (08:45 +0800)]
The found entries by find_get_pages() could be all swap entries.  In
this case we skip the entries, but make sure the skipped entries are
accounted, so we don't keep looping.

Using nr_found > nr_skip to simplify code as suggested by Eric.

Reported-and-tested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/filemap.c

index 645a080..7771871 100644 (file)
@@ -827,13 +827,14 @@ unsigned find_get_pages(struct address_space *mapping, pgoff_t start,
 {
        unsigned int i;
        unsigned int ret;
-       unsigned int nr_found;
+       unsigned int nr_found, nr_skip;
 
        rcu_read_lock();
 restart:
        nr_found = radix_tree_gang_lookup_slot(&mapping->page_tree,
                                (void ***)pages, NULL, start, nr_pages);
        ret = 0;
+       nr_skip = 0;
        for (i = 0; i < nr_found; i++) {
                struct page *page;
 repeat:
@@ -856,6 +857,7 @@ repeat:
                         * here as an exceptional entry: so skip over it -
                         * we only reach this from invalidate_mapping_pages().
                         */
+                       nr_skip++;
                        continue;
                }
 
@@ -876,7 +878,7 @@ repeat:
         * If all entries were removed before we could secure them,
         * try again, because callers stop trying once 0 is returned.
         */
-       if (unlikely(!ret && nr_found))
+       if (unlikely(!ret && nr_found > nr_skip))
                goto restart;
        rcu_read_unlock();
        return ret;