[PATCH] mm: restore vm_normal_page check
authorNick Piggin <npiggin@suse.de>
Sat, 25 Mar 2006 15:20:22 +0000 (16:20 +0100)
committerLinus Torvalds <torvalds@g5.osdl.org>
Sat, 25 Mar 2006 16:43:45 +0000 (08:43 -0800)
Hugh is rightly concerned that the CONFIG_DEBUG_VM coverage has gone too
far in vm_normal_page, considering that we expect production kernels to be
shipped with the option turned off, and that the code has been under some
large changes recently.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
mm/memory.c

index 80c3fb370f91400aae34ae97ce188a9d97629609..e347e106ca3ad4278157431cc39b01ee94bdb931 100644 (file)
@@ -395,12 +395,16 @@ struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, pte_
                        return NULL;
        }
 
-#ifdef CONFIG_DEBUG_VM
+       /*
+        * Add some anal sanity checks for now. Eventually,
+        * we should just do "return pfn_to_page(pfn)", but
+        * in the meantime we check that we get a valid pfn,
+        * and that the resulting page looks ok.
+        */
        if (unlikely(!pfn_valid(pfn))) {
                print_bad_pte(vma, pte, addr);
                return NULL;
        }
-#endif
 
        /*
         * NOTE! We still have PageReserved() pages in the page