Re: [dts] [PATCH] test_plan: add hotplug_test_plan

["Qiu, Michael" <michael.qiu@intel.com>]

On 2/11/2015 10:33 AM, Tetsuya Mukawa wrote:
> On 2015/02/11 1:54, Iremonger, Bernard wrote:
>>> -----Original Message-----
>>> From: Qiu, Michael
>>> Sent: Tuesday, February 10, 2015 8:55 AM
>>> To: Tetsuya Mukawa; dts@dpdk.org
>>> Cc: Iremonger, Bernard
>>> Subject: Re: [PATCH] test_plan: add hotplug_test_plan
>>>
>>> On 2/10/2015 11:24 AM, Tetsuya Mukawa wrote:
>>>> Hi Michael,
>>>>
>>>> On 2015/02/09 22:13, Qiu, Michael wrote:
>>>>> [+cc] Bernard
>>>>>
>>>>> Hi, Bernard and Tetsuya
>>>>>
>>>>> Would you have time to review this test plan?
>>>>> Your comments are very important.
>>>>>
>>>>> Thanks,
>>>>> Michael
>>>>> On 2/2/2015 4:47 PM, Qiu, Michael wrote:
>>> [...]
>>>
>>>>> +
>>>>> +Prerequisites
>>>>> +-------------
>>>>> +Assume DPDK managerd at least one device for physical or none for virtual.
>>>> "managerd" may be typo.
>>>>
>>>>>> +This feature only supports igb_uio now, for uio_pci_generic is on
>>>>>> +the way, will test it after enabled.
>>>>>> +
>>>>>> +To run the testpmd application in linuxapp environment with 4
>>>>>> +lcores,
>>>>>> +4 channels with other default parameters in interactive mode.
>>>>>> +
>>>>>> +        $ ./testpmd -c 0xf -n 4 -- -i
>>>>>> +
>>>>>> +Test ENV:
>>>>>> +
>>>>>> +All test case can be run in 32bit and 64bit platform.
>>>>>> +
>>>>>> +All os can be support, like fedora, ubuntu, rhel, suse, but freebsd
>>>>>> +will not include as hotplug has no plan to support that platform
>>>>>> +
>>>>>> +All kernel version(start from 2.6) can be support, for vfio need kernel
>>>>>> +        version greater than 3.6.
>>>>>> +
>>>>>> +Virtulization support: KVM/VMware/Xen, container is in the roadmap
>>>> "Virtulization" may be typo.
>>>>
>>>>>> +
>>>>>> +-------------------------------------------------------------------
>>>>>> +------------ Test Case 1: port dettach & attach for physical
>>>>>> +devices with igb_uio
>>>>>> +-------------------------------------------------------------------
>>>>>> +------------
>>>>>> +1. run "port attach 0000:02:00.0",
>>>>>> +   assume the physical devices is BDF is 0000:02:00.0
>>>> Above sentence may be typo.
>>> [...]
>>>
>>>> Usually before submitting hotplug patches, I test almost same as
>>>> above, and additionally below cases.
>>>>
>>>> ---------------------------
>>>> Test Re-attach
>>>> ---------------------------
>>>>   - testpmd> attach
>>>>  - testpmd> port start
>>>>  - testpmd> start
>>>>  - testpmd> stop
>>>>  - testpmd> port stop
>>>>  - testpmd> port close
>>>>  - testpmd> port detach
>>>>  - testpmd> port attach
>>>>  - testpmd> port start
>>>>  - testpmd> start
>>>>  - testpmd> stop
>>>>   (To check if re-attached port works fine.)
>>> This will be added for every test case
>>>
>>>> ---------------------------
>>>> start testpmd with "--vdev" option, and detach.
>>>> ---------------------------
>>>>
>>>>   - ./testpmd -c 0xf -n 4 --vdev "eth_pcap0,iface=<interface>" -- -i
>>>>  - testpmd> start
>>>>  - testpmd> stop
>>>>  - testpmd> port stop
>>>>  - testpmd> port close
>>>>  - testpmd> port detach
>>>>    (To make sure if we can detach the port)
>>>>
>>>> ---------------------------
>>>> start testpmd with an attached physical NIC port, and detach.
>>>> ---------------------------
>>>>   - ./dpdk_nic_bind -b igb_uio <pci address>
>>>>  - ./testpmd -c 0xf -n 4 -- -i
>>>>  - testpmd> start
>>>>  - testpmd> stop
>>>>  - testpmd> port stop
>>>>  - testpmd> port close
>>>>  - testpmd> port detach
>>>>    (To make sure if we can detach the port)
>>> I will add both.
>>>
>>> Thanks,
>>> Michael
>>>> How about adding such cases?
>>>>
>>>> Thanks,
>>>> Tetsuya
>>>>
>>>>
>> Hi Michael, Tetsuya,
>>
>> Should there be test cases added using pmd_null (eth_null0) ?
> Hi Bernard,
>
> I guess port hotplug function can be separated in below 2 implementations.
> - port hotplug framework
> - port hotplug implementation in each PMD.
>
> This test plan is for checking hotplug framework.
> So I guess we don't need to add Null pmd cases.

Agree, DTS do not care PMD, it care about different NICs, for DPDK, it
chose PMD for different NICs
 
Thanks,
Michael
> When DTS has tests for each PMD, I guess each PMD tests should have
> cases for hotplug, because each PMD will have different implementation
> for hotplug. I guess null pmd should be also checked like this.
>
> Anyway, if we are not so hurry about adding null pmd tests, I guess I
> will be able to commit it next month.
>
> Thanks,
> Tetsuya
>
>> Regards,
>>
>> Bernard.
>>
>>
>
>