Re: [Qemu-devel] [PATCH] tests: Set read-zeroes on for null-co driver

[Thomas Huth <thuth@redhat.com>]

On 30/07/2019 14.52, Thomas Huth wrote:
> On 29/07/2019 14.46, Andrey Shinkevich wrote:
>> This patch is to reduce the number of Valgrind report messages about
>> using uninitialized memory with the null-co driver. It helps to filter
>> real memory issues and is the same work done for the iotests with the
>> commit ID a6862418fec4072.
>>
>> Suggested-by: Kevin Wolf <kwolf@redhat.com>
>> Signed-off-by: Andrey Shinkevich <andrey.shinkevich@virtuozzo.com>
>> ---
> [...]
>> diff --git a/tests/test-blockjob-txn.c b/tests/test-blockjob-txn.c
>> index 86606f9..7da9216 100644
>> --- a/tests/test-blockjob-txn.c
>> +++ b/tests/test-blockjob-txn.c
>> @@ -15,6 +15,7 @@
>>  #include "qemu/main-loop.h"
>>  #include "block/blockjob_int.h"
>>  #include "sysemu/block-backend.h"
>> +#include "qapi/qmp/qdict.h"
>>  
>>  typedef struct {
>>      BlockJob common;
>> @@ -96,7 +97,9 @@ static BlockJob *test_block_job_start(unsigned int iterations,
>>  
>>      data = g_new0(TestBlockJobCBData, 1);
>>  
>> -    bs = bdrv_open("null-co://", NULL, NULL, 0, &error_abort);
>> +    QDict *opt = qdict_new();
>> +    qdict_put_str(opt, "file.read-zeroes", "on");
>> +    bs = bdrv_open("null-co://", NULL, opt, 0, &error_abort);
>>      g_assert_nonnull(bs);
> 
> Not sure, but don't you need to also qdict_destroy_obj(opt) at the end
> to avoid leaking memory? (Also in the other spots where you use
> qdict_new() ...)

Never mind, seems like bdrv_open() is doing an qobject_unref() on the
options at the end, so I guess this should be fine...

So in that case:

Acked-by: Thomas Huth <thuth@redhat.com>

(I assume this will be taken through the block tree? Let me know if you
prefer the qtest tree instead)