This was triggered, but not the fault of, the dirty page accounting
patches. Suitable for -stable as well, after it goes upstream.
Unable to handle kernel NULL pointer dereference at virtual address
0000004c
EIP is at _spin_lock+0x12/0x66
Call Trace:
[<
401766e7>] __set_page_dirty_buffers+0x15/0xc0
[<
401401e7>] set_page_dirty+0x2c/0x51
[<
40140db2>] set_page_dirty_balance+0xb/0x3b
[<
40145d29>] __do_fault+0x1d8/0x279
[<
40147059>] __handle_mm_fault+0x125/0x951
[<
401133f1>] do_page_fault+0x440/0x59f
[<
4034d0c1>] error_code+0x39/0x40
[<
08048a33>] 0x8048a33
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
*/
int __set_page_dirty_buffers(struct page *page)
{
- struct address_space * const mapping = page->mapping;
+ struct address_space * const mapping = page_mapping(page);
+
+ if (unlikely(!mapping))
+ return !TestSetPageDirty(page);
spin_lock(&mapping->private_lock);
if (page_has_buffers(page)) {