Re: [Qemu-devel] [PATCH 0/2] qcow2: Respect new_block in alloc_refcount_block()

[Max Reitz <mreitz@redhat.com>]

On 2015-02-10 at 17:19, Eric Blake wrote:
> On 02/10/2015 01:02 PM, Max Reitz wrote:
>> Under certain circumstances, making the refcount table grow can result
>> in leaking clusters. The first patch fixes at least some of those
>> circumstances (maybe there are more, but these are the ones I am aware
>> of), and the second patch adds a test case.
>>
>>
>> Max Reitz (2):
>>    qcow2: Respect new_block in alloc_refcount_block()
>>    iotests: Add tests for refcount table growth
> Series:
> Reviewed-by: Eric Blake <eblake@redhat.com>
>
> How'd you find the leak?

Test 015 failed with refcount_bits=32 and refcount_bits=64 after my 
refcount_order series; but strangely only if the recent qcow2 cache 
patch 
(http://lists.nongnu.org/archive/html/qemu-devel/2015-02/msg00786.html) 
was applied before.

Max