find_page:
                page = find_get_page(mapping, index);
                if (!page) {
-                       page_cache_readahead_ondemand(mapping,
-                                       &ra, filp, page,
+                       page_cache_sync_readahead(mapping,
+                                       &ra, filp,
                                        index, last_index - index);
                        page = find_get_page(mapping, index);
                        if (unlikely(page == NULL))
                                goto no_cached_page;
                }
                if (PageReadahead(page)) {
-                       page_cache_readahead_ondemand(mapping,
+                       page_cache_async_readahead(mapping,
                                        &ra, filp, page,
                                        index, last_index - index);
                }
         */
        if (VM_SequentialReadHint(vma)) {
                if (!page) {
-                       page_cache_readahead_ondemand(mapping, ra, file, page,
+                       page_cache_sync_readahead(mapping, ra, file,
                                                           vmf->pgoff, 1);
                        page = find_lock_page(mapping, vmf->pgoff);
                        if (!page)
                                goto no_cached_page;
                }
                if (PageReadahead(page)) {
-                       page_cache_readahead_ondemand(mapping, ra, file, page,
+                       page_cache_async_readahead(mapping, ra, file, page,
                                                           vmf->pgoff, 1);
                }
        }
 
 static unsigned long
 ondemand_readahead(struct address_space *mapping,
                   struct file_ra_state *ra, struct file *filp,
-                  struct page *page, pgoff_t offset,
+                  bool hit_readahead_marker, pgoff_t offset,
                   unsigned long req_size)
 {
        unsigned long max;      /* max readahead pages */
         * Standalone, small read.
         * Read as is, and do not pollute the readahead state.
         */
-       if (!page && !sequential) {
+       if (!hit_readahead_marker && !sequential) {
                return __do_page_cache_readahead(mapping, filp,
                                                offset, req_size, 0);
        }
         * E.g. interleaved reads.
         * Not knowing its readahead pos/size, bet on the minimal possible one.
         */
-       if (page) {
+       if (hit_readahead_marker) {
                ra_index++;
                ra_size = min(4 * ra_size, max);
        }
 }
 
 /**
- * page_cache_readahead_ondemand - generic file readahead
+ * page_cache_sync_readahead - generic file readahead
  * @mapping: address_space which holds the pagecache and I/O vectors
  * @ra: file_ra_state which holds the readahead state
  * @filp: passed on to ->readpage() and ->readpages()
- * @page: the page at @offset, or NULL if non-present
- * @offset: start offset into @mapping, in PAGE_CACHE_SIZE units
+ * @offset: start offset into @mapping, in pagecache page-sized units
  * @req_size: hint: total size of the read which the caller is performing in
- *            PAGE_CACHE_SIZE units
+ *            pagecache pages
  *
- * page_cache_readahead_ondemand() is the entry point of readahead logic.
- * This function should be called when it is time to perform readahead:
- * 1) @page == NULL
- *    A cache miss happened, time for synchronous readahead.
- * 2) @page != NULL && PageReadahead(@page)
- *    A look-ahead hit occured, time for asynchronous readahead.
+ * page_cache_sync_readahead() should be called when a cache miss happened:
+ * it will submit the read.  The readahead logic may decide to piggyback more
+ * pages onto the read request if access patterns suggest it will improve
+ * performance.
  */
-unsigned long
-page_cache_readahead_ondemand(struct address_space *mapping,
-                               struct file_ra_state *ra, struct file *filp,
-                               struct page *page, pgoff_t offset,
-                               unsigned long req_size)
+void page_cache_sync_readahead(struct address_space *mapping,
+                              struct file_ra_state *ra, struct file *filp,
+                              pgoff_t offset, unsigned long req_size)
 {
        /* no read-ahead */
        if (!ra->ra_pages)
-               return 0;
-
-       if (page) {
-               /*
-                * It can be PG_reclaim.
-                */
-               if (PageWriteback(page))
-                       return 0;
-
-               ClearPageReadahead(page);
-
-               /*
-                * Defer asynchronous read-ahead on IO congestion.
-                */
-               if (bdi_read_congested(mapping->backing_dev_info))
-                       return 0;
-       }
+               return;
+
+       /* do read-ahead */
+       ondemand_readahead(mapping, ra, filp, false, offset, req_size);
+}
+EXPORT_SYMBOL_GPL(page_cache_sync_readahead);
+
+/**
+ * page_cache_async_readahead - file readahead for marked pages
+ * @mapping: address_space which holds the pagecache and I/O vectors
+ * @ra: file_ra_state which holds the readahead state
+ * @filp: passed on to ->readpage() and ->readpages()
+ * @page: the page at @offset which has the PG_readahead flag set
+ * @offset: start offset into @mapping, in pagecache page-sized units
+ * @req_size: hint: total size of the read which the caller is performing in
+ *            pagecache pages
+ *
+ * page_cache_async_ondemand() should be called when a page is used which
+ * has the PG_readahead flag: this is a marker to suggest that the application
+ * has used up enough of the readahead window that we should start pulling in
+ * more pages. */
+void
+page_cache_async_readahead(struct address_space *mapping,
+                          struct file_ra_state *ra, struct file *filp,
+                          struct page *page, pgoff_t offset,
+                          unsigned long req_size)
+{
+       /* no read-ahead */
+       if (!ra->ra_pages)
+               return;
+
+       /*
+        * Same bit is used for PG_readahead and PG_reclaim.
+        */
+       if (PageWriteback(page))
+               return;
+
+       ClearPageReadahead(page);
+
+       /*
+        * Defer asynchronous read-ahead on IO congestion.
+        */
+       if (bdi_read_congested(mapping->backing_dev_info))
+               return;
 
        /* do read-ahead */
-       return ondemand_readahead(mapping, ra, filp, page,
-                                       offset, req_size);
+       ondemand_readahead(mapping, ra, filp, true, offset, req_size);
 }
-EXPORT_SYMBOL_GPL(page_cache_readahead_ondemand);
+EXPORT_SYMBOL_GPL(page_cache_async_readahead);