[patch 5/19] grab_cache_page_nowait deadlock fix

From: Andrew Morton (akpm@zip.com.au)
Date: Mon Jun 17 2002 - 01:51:41 EST


- If grab_cache_page_nowait() is to be called while holding a lock on
  a different page, it must perform memory allocations with GFP_NOFS.
  Otherwise it could come back onto the locked page (if it's dirty) and
  deadlock.

  Also tidy this function up a bit - the checks in there were overly
  paranoid.

- In a few of places, look to see if we can avoid a buslocked cycle
  and dirtying of a cacheline.

--- 2.5.22/mm/filemap.c~grab_cache_page_nowait Sun Jun 16 22:50:17 2002
+++ 2.5.22-akpm/mm/filemap.c Sun Jun 16 22:50:17 2002
@@ -445,8 +445,10 @@ int fail_writepage(struct page *page)
 {
         /* Only activate on memory-pressure, not fsync.. */
         if (current->flags & PF_MEMALLOC) {
- activate_page(page);
- SetPageReferenced(page);
+ if (!PageActive(page))
+ activate_page(page);
+ if (!PageReferenced(page))
+ SetPageReferenced(page);
         }
 
         /* Set the page dirty again, unlock */
@@ -868,55 +870,35 @@ struct page *grab_cache_page(struct addr
  * This is intended for speculative data generators, where the data can
  * be regenerated if the page couldn't be grabbed. This routine should
  * be safe to call while holding the lock for another page.
+ *
+ * Clear __GFP_FS when allocating the page to avoid recursion into the fs
+ * and deadlock against the caller's locked page.
  */
-struct page *grab_cache_page_nowait(struct address_space *mapping, unsigned long index)
+struct page *
+grab_cache_page_nowait(struct address_space *mapping, unsigned long index)
 {
- struct page *page;
-
- page = find_get_page(mapping, index);
-
- if ( page ) {
- if ( !TestSetPageLocked(page) ) {
- /* Page found and locked */
- /* This test is overly paranoid, but what the heck... */
- if ( unlikely(page->mapping != mapping || page->index != index) ) {
- /* Someone reallocated this page under us. */
- unlock_page(page);
- page_cache_release(page);
- return NULL;
- } else {
- return page;
- }
- } else {
- /* Page locked by someone else */
- page_cache_release(page);
- return NULL;
- }
- }
-
- page = page_cache_alloc(mapping);
- if (unlikely(!page))
- return NULL; /* Failed to allocate a page */
+ struct page *page = find_get_page(mapping, index);
 
- if (unlikely(add_to_page_cache_unique(page, mapping, index))) {
- /*
- * Someone else grabbed the page already, or
- * failed to allocate a radix-tree node
- */
+ if (page) {
+ if (!TestSetPageLocked(page))
+ return page;
                 page_cache_release(page);
                 return NULL;
         }
-
+ page = alloc_pages(mapping->gfp_mask & ~__GFP_FS, 0);
+ if (page && add_to_page_cache_unique(page, mapping, index)) {
+ page_cache_release(page);
+ page = NULL;
+ }
         return page;
 }
 
 /*
  * Mark a page as having seen activity.
  *
- * If it was already so marked, move it
- * to the active queue and drop the referenced
- * bit. Otherwise, just mark it for future
- * action..
+ * inactive,unreferenced -> inactive,referenced
+ * inactive,referenced -> active,unreferenced
+ * active,unreferenced -> active,referenced
  */
 void mark_page_accessed(struct page *page)
 {
@@ -924,10 +906,9 @@ void mark_page_accessed(struct page *pag
                 activate_page(page);
                 ClearPageReferenced(page);
                 return;
+ } else if (!PageReferenced(page)) {
+ SetPageReferenced(page);
         }
-
- /* Mark the page referenced, AFTER checking for previous usage.. */
- SetPageReferenced(page);
 }
 
 /*
@@ -2286,7 +2267,8 @@ generic_file_write(struct file *file, co
                         }
                 }
                 kunmap(page);
- SetPageReferenced(page);
+ if (!PageReferenced(page))
+ SetPageReferenced(page);
                 unlock_page(page);
                 page_cache_release(page);
                 if (status < 0)

-
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Sun Jun 23 2002 - 22:00:12 EST