From: Hirokazu Takahashi Hi Dave, The following patch makes unwind() call unlock(newpage) earlier. This removes the bad restriction that any kernel code must call put_page() before lock_page() when the code notices unwind() has been called. I realized lock_replace_page() in the previous patch required this patch. Please apply it after P10-rename_rewind_page_to_unwind_page.patch . Thanks, Hirokazu Takahashi. Signed-off-by: Dave Hansen --- memhotplug-dave/mm/mmigrate.c | 2 +- 1 files changed, 1 insertion(+), 1 deletion(-) diff -puN mm/mmigrate.c~AA-PM-10.2-fix-rewind-locking mm/mmigrate.c --- memhotplug/mm/mmigrate.c~AA-PM-10.2-fix-rewind-locking 2005-03-28 16:38:16.000000000 -0800 +++ memhotplug-dave/mm/mmigrate.c 2005-03-28 16:38:16.000000000 -0800 @@ -202,6 +202,7 @@ out: ClearPageSwapCache(newpage); /* XXX unmap needed? No, it shouldn't. Handled by fault handlers. */ unlock_page(newpage); + unlock_page(page); /* * Some requests may be blocked on the newpage. Wait until the @@ -224,7 +225,6 @@ out: BUG_ON(PageActive(newpage)); BUG_ON(PagePrivate(newpage)); BUG_ON(page_count(newpage) != 2); - unlock_page(page); page_cache_release(newpage); return 0; } _