On 06/30/2016 10:00 AM, Jeff Cody wrote:
> On Thu, Jun 30, 2016 at 04:45:52PM +0800, Changlong Xie wrote:
>> On 06/30/2016 04:25 PM, Fam Zheng wrote:
>>> On Thu, 06/30 16:01, Changlong Xie wrote:
>>>> Otherwise, we could never trigger assert(!bitmap->successor)
>>>>
>>>> Signed-off-by: Changlong Xie <xiecl.f...@cn.fujitsu.com>
>>>> ---
>>>> block/dirty-bitmap.c | 1 -
>>>> 1 file changed, 1 deletion(-)
>>>>
>>>> diff --git a/block/dirty-bitmap.c b/block/dirty-bitmap.c
>>>> index 4902ca5..e9df5ac 100644
>>>> --- a/block/dirty-bitmap.c
>>>> +++ b/block/dirty-bitmap.c
>>>> @@ -131,7 +131,6 @@ int
>>>> bdrv_dirty_bitmap_create_successor(BlockDriverState *bs,
>>>> if (bdrv_dirty_bitmap_frozen(bitmap)) {
>>>> error_setg(errp, "Cannot create a successor for a bitmap that is "
>>>> "currently frozen");
>>>> - return -1;
>>>> }
>>>> assert(!bitmap->successor);
>>>
>>> This is wrong. Then we will always trigger assert for a frozen bitmap.
>>>
>>
>> IMO, when it's a frozen bitmap, we will always return -1. So
>> "assert(!bitmap->successor)" is useless here, am i right?
>>
>
> I don't see a path where the assert could trigger, so I would agree that the
> assert itself, while harmless, is not necessary (although it could be argued
> it is in place in case the code above it changes in a way that does not
> check bitmap->successor).
>
> That doesn't mean we want to try and trigger an assert, however! :) The
> error return is the proper error handling -- we don't expect that asserts
> should ever be encountered QEMU, if one happens that is a sign of a bug.
>
> Jeff
>
The assert was indeed added to ensure that if the valid states of the
bitmap later expanded or changed, that the status checkers (e.g.
bdrv_dirty_bitmap_frozen()) were changed to match.
