[dpdk-dev] [PATCH 2/7] net/qede: refactor filtering code

Harish Patil harish.patil at qlogic.com
Wed Dec 21 00:12:12 CET 2016


>
>On 12/12/2016 5:36 PM, Harish Patil wrote:
>> 
>>> On 12/3/2016 2:43 AM, Harish Patil wrote:
>>>> The filter_config in qed_eth_ops_pass is a wrapper call driving all
>>>>the
>>>> filtering configuration. This requires defining multiple structures
>>>>and
>>>> passing different function arguments for each filter type which is
>>>> unnecessary. So as part of this refactoring we remove filter_config
>>>>from
>>>> qed_eth_ops_pass and invoke base apis directly. Another change is to
>>>> use a singly list for unicast/multicast macs and thereby prevent
>>>> duplicate
>>>
>>> singly linked list?
>> 
>> Yes. Two lists to track unicast and multicast mac entries:
>> +       SLIST_HEAD(mc_list_head, qede_mcast_entry) mc_list_head;
>> 
>> +       SLIST_HEAD(uc_list_head, qede_ucast_entry) uc_list_head;
>> 
>> 
>> 
>>>
>>>> entries.
>>>>
>>>> This change is primarily intended to support future tunneling support
>>>> which makes use of existing L2 filtering/classifications.
>>>>
>>>> Fixes: 2ea6f76a ("qede: add core driver")
>>>
>>> What is fixed in this patch, isn't it a refactor?
>> 
>> The fix part is to make use of the newly added lists and prevent
>>duplicate
>> mac filters.
>> Before that there were no checks. Other than that its refactoring to
>> mainly invoke direct base APIs.
>
>So this is not fixing any defect in driver, so I believe Fixes tag can
>be removed. This tag is mainly useful to pick commits for stable trees.

It also fixes the driver. Apart from preventing duplicate filter
programming, the existing qede_mac_addr_add() did not differentiate
between multicast and unicast mac entries. So the multicast mac addition
would wrongly go to unicast table.
So we would need Fixes tag and would need to be picked up for stable trees.

>
>> 
>>>
>>> btw, all Fixes formats are wrong in the patchset, can you please use
>>>the
>>> git alias provided:
>>> 
>>>http://dpdk.org/doc/guides/contributing/patches.html#commit-messages-bod
>>>y

OK sure.

>>>
>>>>
>>>> Signed-off-by: Harish Patil <harish.patil at qlogic.com>
>>>> ---
>>> <...>
>>>
>> 
>> 
>
>




More information about the dev mailing list