Hi Ferruh,
On 4/17/20 5:44 PM, Ferruh Yigit wrote:
> On 4/17/2020 4:14 PM, Maxime Coquelin wrote:
>> Hi Ferruh & Andrew,
>>
>> On 4/17/20 8:40 AM, Ivan Dyukov wrote:
>>> Hello Everyone,
>>>
>>> Ook. I can take care about examples updates. link_speed could be printed
>>> in following way:
>>> ("link speed %u%s", link_speed, link_speed
>>> ==ETH_SPEED_NUM_UNKNOWN?"(UNKNOWN)":"")
>>>
>>> Please let me know if you have any objections.
>>>
>>> There are about 47 cases.
>>>
>>> $ grep -rn link_speed examples/ app/ doc/ | wc -l
>>> 47
>>>
>>> Thanks,
>>> Ivan
>>> 17.04.2020 01:14, Thomas Monjalon пишет:
>>>> Hi,
>>>>
>>>> Please look at an update below from ethdev co-maintainers.
>>>>
>>>> 16/04/2020 14:42, Ivan Dyukov:
>>>>> UNKNOWN speed equals to 0xffffffff
>>>> [...]
>>>>> +#define ETH_SPEED_NUM_UNKNOWN 0xffffffff /**< Unknown */
>>>> This approach is being rejected in another thread:
>>>> https://protect2.fireeye.com/url?k=ed2d2a70-b0fe73ce-ed2ca13f-0cc47a31ba82-80584d32127c24cd&q=1&u=http%3A%2F%2Finbox.dpdk.org%2Fdev%2F42de4bd1-0a6c-6591-cd27-67ce692fabc9%40intel.com%2F
>>
>> Would that work for you?
>> I would need your ACK before applying the series (which I planned to do
>> for -rc1).
>
> Hi Maxime,
>
> There is another patch from Thomas that targets this change only [1], and it
> is
> waiting for change request, because the scope of the change is larger than
> just
> defining a new macro, documentation & sample/test applications should be aware
> of this new speed definition.
>
> Instead of this patch, this patchset can wait [1] as dependency.
>
> Or if this patchset is urgent, perhaps this patch can go in as it as and
> Thomas'
> patch can replace it later with full implementation, if Thomas agrees.
>
> And not sure if it is good idea, but perhaps this "unknown speed' can be used
> local to virtio until [1] becomes ready, though I am for having this as last
> option.
I was replying to my mail after discussing with Thomas.
I agree Thomas's series is better, and now understand this is an API
change that was not announced.
What I propose it basically to apply Ivan's v8 with a few fixes on top
that he did. It means that if no value is set in the NIC or not value
defined as devargs, then 10G will be picked as default.
Note that the 10G value is the one currently displayyed, without Ivan's
series, so it seems a reasonable temporary solution.
Thanks,
Maxime
> [1]
> https://patches.dpdk.org/patch/67915/
>
