Re: [PATCH] KVM: fix the handling of dirty bitmaps to avoid overflows

(2010/04/13 2:39), Marcelo Tosatti wrote:
> On Mon, Apr 12, 2010 at 07:35:35PM +0900, Takuya Yoshikawa wrote:
>> This patch fixes a bug found by Avi during the review process
>> of my dirty bitmap related work.
>>
>> To ppc and ia64 people:
>>    The fix is really simple but touches all architectures using
>>    dirty bitmaps. So please check this will not suffer your part.
>>
>> =>>
>> Int is not long enough to store the size of a dirty bitmap.
>>
>> This patch fixes this problem with the introduction of a wrapper
>> function to calculate the sizes of dirty bitmaps.
>>
>> Note: in mark_page_dirty(), we have to consider the fact that
>>    __set_bit() takes the offset as int, not long.
>>
>> Signed-off-by: Takuya Yoshikawa<[email protected]>
>
> Applied, thanks.
>

Thanks everyone!

BTW, just from my curiosity, are there any cases in which we use such huge
number of pages currently?

   ALIGN(memslot->npages, BITS_PER_LONG) / 8;

More than G pages need really big memory!
   -- We are assuming some special cases like "short" int size?


If so, we may have to care about a lot of things from now on, because common
functions like __set_bit() don't support such long buffers.

If not, my patch might be over hacking -- especially the following part:


@@ -1183,10 +1183,13 @@ void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
  	memslot = gfn_to_memslot_unaliased(kvm, gfn);
  	if (memslot && memslot->dirty_bitmap) {
  		unsigned long rel_gfn = gfn - memslot->base_gfn;
+		unsigned long *p = memslot->dirty_bitmap +
+					rel_gfn / BITS_PER_LONG;
+		int offset = rel_gfn % BITS_PER_LONG;

  		/* avoid RMW */
-		if (!generic_test_le_bit(rel_gfn, memslot->dirty_bitmap))
-			generic___set_le_bit(rel_gfn, memslot->dirty_bitmap);
+		if (!generic_test_le_bit(offset, p))
+			generic___set_le_bit(offset, p);
  	}
  }