Re: [PATCH 4/7] iotests/129: Use throttle node

[Max Reitz <mreitz@redhat.com>]

On 13.01.21 16:07, Vladimir Sementsov-Ogievskiy wrote:
> 13.01.2021 17:10, Max Reitz wrote:
>> On 13.01.21 15:06, Max Reitz wrote:
>>> Throttling on the BB has not affected block jobs in a while, so it is
>>> possible that one of the jobs in 129 finishes before the VM is stopped.
>>> We can fix that by running the job from a throttle node.
>>>
>>> Signed-off-by: Max Reitz <mreitz@redhat.com>
>>> ---
>>>   tests/qemu-iotests/129 | 39 +++++++++++++++------------------------
>>>   1 file changed, 15 insertions(+), 24 deletions(-)
>>>
>>> diff --git a/tests/qemu-iotests/129 b/tests/qemu-iotests/129
>>> index dd23bb2e5a..febc806398 100755
>>> --- a/tests/qemu-iotests/129
>>> +++ b/tests/qemu-iotests/129
>>
>> [...]
>>
>>> @@ -53,33 +51,26 @@ class TestStopWithBlockJob(iotests.QMPTestCase):
>>>       def do_test_stop(self, cmd, **args):
>>>           """Test 'stop' while block job is running on a throttled 
>>> drive.
>>>           The 'stop' command shouldn't drain the job"""
>>> -        params = {"device": "drive0",
>>> -                  "bps": 1024,
>>> -                  "bps_rd": 0,
>>> -                  "bps_wr": 0,
>>> -                  "iops": 0,
>>> -                  "iops_rd": 0,
>>> -                  "iops_wr": 0,
>>> -                 }
>>> -        result = self.vm.qmp("block_set_io_throttle", conv_keys=False,
>>> -                             **params)
>>> -        self.assert_qmp(result, 'return', {})
>>>           result = self.vm.qmp(cmd, **args)
>>>           self.assert_qmp(result, 'return', {})
>>> +
>>>           result = self.vm.qmp("stop")
>>>           self.assert_qmp(result, 'return', {})
>>>           result = self.vm.qmp("query-block-jobs")
>>> +
>>>           self.assert_qmp(result, 'return[0]/status', 'running')
>>>           self.assert_qmp(result, 'return[0]/ready', False)
>>> +        self.vm.qmp("block-job-cancel", device="drive0", force=True)
>>> +
>>
>> Sorry, somewhere along the way I forgot to remove this 
>> block-job-cancel.  It’s still from
>> https://lists.nongnu.org/archive/html/qemu-block/2019-06/msg00499.html 
>> and I don’t know why I added it there.  It should probably go.
>>
>> ('stop' should have drained, that’s the point, so there shouldn’t be 
>> any further delay if we quit the VM without cancelling the job.)
>>
> 
> what do you mean by 'stop' should have drained? As I understand, after 
> "stop" jobs should continue and it's OK.. So, immediately after "stop" 
> command job may generate new requests, why not?

Hm, right.

Well, either way, the cancel shouldn’t make a difference.  If quitting 
the VM were to stall because of the job, then cancelling it beforehand 
only means that then cancel will stall.

Max