[dpdk-dev] [RFC V4] app/testpmd: fix queue stats mapping configuration

[Min Hu (Connor)]

Hi, Ferruh, and all,
	are there any comments about this set of patch?

在 2020/11/23 17:51, Ferruh Yigit 写道:
> On 11/21/2020 4:29 AM, Stephen Hemminger wrote:
>> On Fri, 20 Nov 2020 23:33:40 +0000
>> Ferruh Yigit <ferruh.yigit at intel.com> wrote:
>>
>>> On 11/20/2020 11:21 PM, Stephen Hemminger wrote:
>>>> On Fri, 20 Nov 2020 17:26:55 +0000
>>>> Ferruh Yigit <ferruh.yigit at intel.com> wrote:
>>>>> On 11/20/2020 11:50 AM, Min Hu (Connor) wrote:
>>>>>> From: Huisong Li <lihuisong at huawei.com>
>>>>>>
>>>>>> Currently, the queue stats mapping has the following problems:
>>>>>> 1) Many PMD drivers don't support queue stats mapping. But there 
>>>>>> is no
>>>>>>     failure message after executing the command "set stat_qmap rx 
>>>>>> 0 2 2".
>>>>>> 2) Once queue mapping is set, unrelated and unmapped queues are also
>>>>>> displayed.
>>>>>> 3) The configuration result does not take effect or can not be 
>>>>>> queried
>>>>>> in real time.
>>>>>> 4) The mapping arrays, "tx_queue_stats_mappings_array" &
>>>>>> "rx_queue_stats_mappings_array" are global and their sizes are 
>>>>>> based on
>>>>>> fixed max port and queue size assumptions.
>>>>>> 5) These record structures, 
>>>>>> 'map_port_queue_stats_mapping_registers()'
>>>>>>     and its sub functions are redundant for majority of drivers.
>>>>>> 6) The display of the queue stats and queue stats mapping is mixed
>>>>>>     together.
>>>>>>
>>>>>> Since xstats is used to obtain queue statistics, we have made the 
>>>>>> following
>>>>>>     simplifications and adjustments:
>>>>>> 1) If PMD requires and supports queue stats mapping, configure to 
>>>>>> driver in
>>>>>>     real time by calling ethdev API after executing the command
>>>>>>     "set stat_qmap rx/tx ...". If not, the command can not be 
>>>>>> accepted.
>>>>>> 2) Based on the above adjustments, these record structures,
>>>>>> 'map_port_queue_stats_mapping_registers()' and its sub functions 
>>>>>> can be
>>>>>>     removed. "tx-queue-stats-mapping" & "rx-queue-stats-mapping" 
>>>>>> parameters,
>>>>>>     and 'parse_queue_stats_mapping_config()' can be removed too.
>>>>>> 3) remove display of queue stats mapping in 'fwd_stats_display()' &
>>>>>>     'nic_stats_display()', and obtain queue stats by xstats.
>>>>>>     Since the record structures are removed, 
>>>>>> 'nic_stats_mapping_display()'
>>>>>>     can be deleted.
>>>>>>
>>>>>> Fixes: 4dccdc789bf4b ("app/testpmd: simplify handling of stats 
>>>>>> mappings error")
>>>>>> Fixes: 013af9b6b64f6 ("app/testpmd: various updates")
>>>>>> Fixes: ed30d9b691b21 ("app/testpmd: add stats per queue")
>>>>>>
>>>>>> Signed-off-by: Huisong Li <lihuisong at huawei.com>
>>>>>
>>>>> Overall    looks good to me.
>>>>> I did a quick test didn't see anything unexpected. 'xstats' or 
>>>>> 'dpdk-proc-info'
>>>>> app still can be used to get queue stats and "set stat_qmap ..." is 
>>>>> working as
>>>>> expected.
>>>>>
>>>>> But it is a little late for this release cycle, would you be OK to 
>>>>> get this at
>>>>> the beginning of next release?
>>>>
>>>> Could we plan to deprecate queue stats mapping in future when xstats 
>>>> work is done?
>>>
>>> Even queue stats moved to xstats, a few PMDs still need this 
>>> configuration and API.
>>> And this patch already cleans the queue stats mapping noise from 
>>> testpmd.
>>>
>>> What is the benefit/motivation to deprecate the queue stats mapping API?
>>
>> Mostly because so few drivers implement it, that any application using it
>> is  going to be non-portable.
>>
> 
> 'rte_eth_dev_set_.*_queue_stats_mapping()' APIs return '-ENOTSUP' if 
> underlying PMD doesn't support it, applications can check this for 
> portable code.
> .